Calvin Huang · Patient C · multi-platform biology map · MDT

v0.2 · 2026-05-12

Calvin Huang · DHG H3.3 G34R · WHO 4 · LMD · MDT decision-support

Biology map · pre-therapy

An integrated view of Calvin Huang's tumour across every assay platform, transcriptomic (short-read and long-read mRNA), proteomic (mProbe MRM-MS, Ignite RPPA), genomic (tissue WES, CSF ctDNA, fusions, CNV, germline/MMR), and ex-vivo drug response, with each finding shown alongside its cross-platform corroboration. The goal is to present the full state of the molecular information on this patient, cross-referenced, for multidisciplinary treatment decisions.

PatientCalvin Huang · age 29

DiagnosisH3.3 G34R diffuse hemispheric glioma · WHO grade 4

MolecularH3-3A G34R · TP53 R273C · ATRX-loss / ALT

Disease stateLeptomeningeal dissemination

First DxFeb 2016

Primary siteRight frontal / hemispheric

Current specimenSSW-24-26425 · 15 Oct 2024

Latest imaging30 Apr / 1 May 2026, mixed, not RANO PR

Current therapyreconcile vs drug diary

Caris short-readReported 2025-08-02N = 19,168 genes (TPM)

Long-read (UCLA tissue)Reported 2024-12-23N = 19,952 genes (CPM)

vs GTEx normal brainn = 17 GTEx controlsACC · FC · CH · Putamen

mProbe MRM-MS2024 specimen · reported 2026-04-26282-protein targeted

2016 GPS CancerNantOmics WGS/RNA + proteomicsbaseline comparator

Calvin Huang, integrated overview

Calvin Huang, 29-year-old man. H3.3 G34R diffuse hemispheric glioma, WHO grade 4, with leptomeningeal dissemination. Diagnosed February 2016. Two resections; the current reference specimen is the 15 Oct 2024 UCLA Liau resection. Prior therapy: chemoradiotherapy with temozolomide, then lomustine, ONC201/206, anti-angiogenic and immunotherapy lines, and CSF-directed cellular therapy; craniospinal proton irradiation completed 8 May 2026. Most recent imaging (30 Apr / 1 May 2026 MRI): first non-progression since November 2025, mixed / stabilisation, not a formal RANO partial response. Active and recent therapy to be reconciled against the patient’s drug diary.

Data layers (all on the 15 Oct 2024 specimen unless stated): Caris short-read mRNA (N=19,168 filtered); UCLA long-read mRNA (N=19,952 filtered, isoform-aware); mProbe MRM-MS targeted proteomics (282 proteins; 2016 GPS baseline); Ignite RPPA phospho-protein panel (35 analytes, expanded 15 May 2026); Filbin GBX63 PDX ex-vivo drug response; Caris WES somatic DNA; two CSF ctDNA programmes, SiMSen-Seq personalised weekly kinetics, and the multi-platform Belay / Shanghai Decode / OncoIncytes series.

Four sequentially dominant populations · 2016 → 2026

2016 founding

Bulk drug-sensitive

Partial resection MGH + Stupp (60 Gy + TMZ ×6) + DC-Vax. H3.3 G35R + TP53 R273C only. TMB 1.0. Eight years of stable disease. ERCC1 / hENT1 / RRM1 / TOPO1 proteins all detected. CDK6 not amplified.

Oct 2024 recurrence

Treatment-remodelled

UCLA Liau resection (7 specimens incl. falx, dural involvement at Oct 2024, 15 months before formal LMD diagnosis). CDK6 amplification acquired. ATRX lost. ERCC1/hENT1/RRM1/MSH2/MSH6 proteins all lost. TMB 1.0 → 4.31.

Jul 2025 CSF

DDR subclone (cleared)

Predicine CARE CSF: RAD50 fs + ATM H1951R + BAP1 H141R + BRCA1 Q1556R at 0.6–2.5% VAF. Compounded repair deficiency, most chemosensitive subclone. Selectively eliminated by Feb 2026.

May 2026 current

LMD residual population

SiMSen-Seq personalised CSF: TP53 R273C 51.16% VAF · 2,095 MM/mL · 16 patient-specific variants. 30 Apr / 1 May 2026 UMiami MRI: first non-progression since Nov 2025 (mixed / stabilisation mid-CSI, not a formal RANO partial response per source MDT & imaging docs; the 12 Apr scan was the worst to date). On CSI proton + ONC201/206/2-DG + Ad-TD-nsIL12. Grade 4 lymphopenia (ALC 0.21). Prior CAR-T ×4 (all antigens) progressed through every cycle.

Biology axes

Each card is a multi-platform-supported biological surface a therapy plan must address, the hypothesis-led axes plus an unbiased whole-transcriptome outlier scan and the integrative NF-κB / XPO1 axis. Click any card to drill into its per-gene data.

Axis 01

CDK6 / cell-cycle

Acquired CDK6 amplification + chr7q amplicon-associated rearrangements (CDK6:VIPR2 292 reads, corroborate amplification, not oncogenic fusions). Liu 2024 lineage-specific vulnerability. Filbin ex-vivo CDK4/6i validated. CCND2 Z +3.76, CDK4 Z +3.16 reinforce cyclin-D drive. Ignite RPPA: RB pS780 only 14% + FOXM1 pT600 4%, RB phospho-substrate is low, sharpening the RB-loss durability concern.

Trunk · always on

Axis 02

DNA repair catastrophe

HR + NHEJ + MMR + NER effector arms ND at protein (MRM-MS, panel caveats apply). ATRX-loss (IHC). PARP1 robust (1,863 amol/μg). Probable germline CDK12 P530A. Filbin niraparib ex-vivo 0.06. Ignite RPPA expanded panel: CHK2 pS33/S35 99%, CHK1 pS345 95%, the two highest activations measured; DNA-damage checkpoint maximally engaged → checkpoint-inhibition rationale (CHK1/2i · ATRi · WEE1i).

Trunk + pulse

Axis 03

Ferroptosis-vulnerable state on R273C lineage

TP53 R273C truncal DNA-contact mutant, 51.16% SiMSen CSF VAF. AIFM2 (FSP1) Z −0.94 (19th-percentile, GSH-independent rescue reduced); GPX4 retained, NRF2 antioxidant programme high → GSH-dependent state. Three agents converge on the same arm: APR-246, sulfasalazine, auranofin. HSP90 abundant (13,792 amol/μg), chaperone destabilisation a separate exploratory route.

Pulse archetype

Axis 04

Multi-RTK bypass

PDGFRA Z +3.87 (plus PDGFRA p.Y288C activating in CSF ctDNA, recurrence-emergent) · FGFR1/2/3 all Z >3 · NTRK2 Z +4.26 · EGFR Z +3.3, six RTK nodes parallel-expressed and Filbin ex-vivo validated. Ignite RPPA activation states mostly low (FGFR 10%, PDGFRB 7%, TRK 13%, EGFR 7%, RET 9%, ALK 13%) but MET phospho 20%, supports MET-active interpretation already seen on Schürch IHC. Filbin synergies: ribo + avap 5.68 · + ponatinib 5.91 · + bumetanide 5.67.

Rotating

Axis 05

Vertical cascade

AKT3 Z +4.71 · PIK3R1 Z +4.67 · mTOR/TSC1/TSC2 present · PTEN Z +3.88. GSEA axis-suppressed. Ignite RPPA confirms PI3K/AKT/mTOR not constitutively active (AKT pT308 14%, pS473 6%, mTOR 22%, p70S6K 12%, S6RP 5%, 4EBP1 7%). ERK1/2 pT202/Y204 89% with MEK 12%, and the expanded panel adds BRAF pS445 52%: upstream MAPK is more active than the MEK site alone implied. STAT5 pY694 17% low.

Rotating

Axis 06

G34R lineage + chromatin

Long-read logFC FOXG1 +13.67 · DLX1 +9.91 · ASCL1 +4.5. Super-enhancer-driven GABAergic interneuron precursor state confirmed across all 4 GTEx normal-brain comparisons. Chromatin-memory window 4-8 wks (Oliveira 2025).

Continuous + pulse

Axis 07

Cold-but-rescuable immune

B2M logFC +8.88 vs GTEx · MHC-I hyperinduced. CXCL9 / CXCL10 = 0 · IFNG 0.35 near-absent. CSF1R Z +3.44 TAM-dominant. CD47 protein 2,761. Ignite RPPA PD-L1 58%, contradicts mProbe ND and Caris mRNA low; functional PD-L1 surface signal moderate-high. Combination required given cold TME (CXCL9/10 induction + checkpoint).

Continuous + episodic

Axis 08

Treatment-survivor metabolism

2016→2024: ERCC1 / hENT1 / RRM1 / TOPO1 all lost at protein. PKM 25,706 amol (glycolysis dominant). GNAS+PRKAR1A cAMP axis Z +5.29/+4.99. Carboplatin never tried despite ERCC1-ND for a decade.

Continuous + pulse

Axis 09

Cross-platform outliers

31 genes that rank top-decile on both Caris and long-read but sit outside the hypothesis-led axes. CLU (Z +4.47 / +3.83), pro-survival chaperone, TKI-resistance factor, is the highest-ranked actionable candidate in this set. Lineage layer (GFAP / AQP4 / S100B / SLC1A2 / NDRG2) plus motility layer (YWHAE / RAC1 / TMSB4X-10) plus chaperone layer (CLU / HSPA8).

Discovery scan

Axis 10

NF-κB / nuclear export (XPO1–IκBα)

NFKBIA (IκBα) over-expressed concordantly on both RNA platforms (logFC vs brain +1.10), with NRF2-high and RAN-high, a coherent nuclear-export / NF-κB-survival signature. XPO1 exports IκBα; selinexor reactivates it. The patient’s 4-line screen is selinexor-sensitive. Selinexor is the act-on readout (experimental).

Experimental node

Reading the multi-platform tables

Z-score interpretation

Z = (log₁₀(TPM+1) − mean) / SD across all biology-curated genes in the dataset (mito + pseudogenes excluded). Z > +2 = top ~2% of expressed genome. Z > +3 = top ~0.3%. Z > +4 = top ~0.01%. Negative Z = below mean expression.

Why both platforms?

Caris short-read = exome-aware, deeper sampling for low-expressed genes. Long-read = isoform/fusion-aware, better at structural events but shallower coverage. Agreement across both (Z > +2 on both) is the most reliable single indicator of biological relevance.

logFC vs GTEx

log₂(tumour CPM / median GTEx CPM) across 17 normal-brain controls (ACC, FC, CH, Putamen). Captures tumour-specific over-expression independent of housekeeping baseline. On the log₂ scale, logFC +1 ≈ 2×, +2 ≈ 4×, and +3 ≈ 8× expression over normal brain.

RNA / transcriptome, summary

Consolidated transcriptome across the two RNA platforms. Caris short-read TPM is ranked within the Caris pan-tumour FFPE cohort; UCLA long-read CPM is compared to 17 GTEx normal-brain controls (logFC).

Cross-platform concordance

Genome-wide Caris-vs-long-read Z correlation is weak (Pearson r = 0.38). Of 663 genes Caris ranks as high outliers (Z > +2), 273 (41%) are not corroborated by long-read. Neither platform is ground truth (see below): short-read rank is precise within a pan-tumour cohort but carries systematic bias for homologous families; long-read undercounts and the long-read-vs-GTEx contrast is itself cross-platform. Act-on calls require cross-platform concordance plus orthogonal or functional support; single-platform calls are hypothesis-only. For clonal questions, absolute CSF MM/mL is the metric.

Concordant signals

Gene	Caris %ile	Caris Z	LR Z	logFC vs brain	Read
CDK6	96	+1.98	+3.08	+6.15	Cross-platform concordant; ex-vivo CDK4/6i active — act-on
CCND2	100	+3.76	+2.48	+2.50	Cross-platform concordant; reinforces cyclin-D drive
NFKBIA (IκBα)	100	+3.55	+2.02	+1.10	Cross-platform concordant; NF-κB / XPO1-cargo (selinexor-relevant)
TUBB3	94	+1.65	+2.19	+1.45	Cross-platform concordant; taxane/vinca-resistance biology
NFE2L2 (NRF2)	99	+3.10	+1.46	+1.90	Cross-platform concordant; antioxidant-survival context

Single-platform calls

High in the Caris pan-tumour cohort but not concordant across platforms.

Gene	Caris %ile	Caris Z	LR Z	logFC vs brain	Read
PDGFRA	100	+3.87	+0.80	−1.72	Top-percentile in the Caris pan-tumour cohort but not corroborated by long-read or tumour-vs-brain; single-platform — hypothesis-only, needs orthogonal / functional confirmation
PDGFRB	99	+3.03	−0.19	−1.37	Top-percentile in the Caris pan-tumour cohort but not corroborated by long-read or tumour-vs-brain; single-platform — hypothesis-only, needs orthogonal / functional confirmation
KIT	98	+2.30	−0.45	−2.07	Top-percentile in the Caris pan-tumour cohort but not corroborated by long-read or tumour-vs-brain; single-platform — hypothesis-only, needs orthogonal / functional confirmation
AKT3	100	+4.71	−0.67	−4.22	Top-percentile in the Caris pan-tumour cohort but not corroborated by long-read or tumour-vs-brain; single-platform — hypothesis-only, needs orthogonal / functional confirmation
MAPK1	100	+3.55	+0.20	−5.54	Top-percentile in the Caris pan-tumour cohort but not corroborated by long-read or tumour-vs-brain; single-platform — hypothesis-only, needs orthogonal / functional confirmation
MAP2K1	99	+2.97	+0.11	−6.84	Top-percentile in the Caris pan-tumour cohort but not corroborated by long-read or tumour-vs-brain; single-platform — hypothesis-only, needs orthogonal / functional confirmation
PIK3CA	98	+2.58	−0.92	−2.28	Top-percentile in the Caris pan-tumour cohort but not corroborated by long-read or tumour-vs-brain; single-platform — hypothesis-only, needs orthogonal / functional confirmation
EGFR	99	+3.24	−0.05	+0.07	Top-percentile in the Caris pan-tumour cohort but not corroborated by long-read or tumour-vs-brain; single-platform — hypothesis-only, needs orthogonal / functional confirmation
FGFR1	100	+3.55	−0.62	−1.83	Top-percentile in the Caris pan-tumour cohort but not corroborated by long-read or tumour-vs-brain; single-platform — hypothesis-only, needs orthogonal / functional confirmation
FGFR2	100	+3.62	−0.12	−3.84	Top-percentile in the Caris pan-tumour cohort but not corroborated by long-read or tumour-vs-brain; single-platform — hypothesis-only, needs orthogonal / functional confirmation
NTRK3	99	+3.29	−0.16	−3.45	Top-percentile in the Caris pan-tumour cohort but not corroborated by long-read or tumour-vs-brain; single-platform — hypothesis-only, needs orthogonal / functional confirmation
TACC3	97	+2.25	−1.05	−0.66	Top-percentile in the Caris pan-tumour cohort but not corroborated by long-read or tumour-vs-brain; single-platform — hypothesis-only, needs orthogonal / functional confirmation
NTRK2	100	+4.26	+2.27	+2.23	Concordant on long-read, but NTRK is actionable only as a fusion (none present); high TrkB is neural-lineage expression

Why no single platform is the arbiter

Caris short-read (internal cohort)

The Caris percentile is a rank within Caris's pan-tumour FFPE reference of tens of thousands of cases. The rank itself is statistically precise (large n, low random error), and the genes above sit at the 94–100th percentile there — a real signal in that frame. The reference is pan-tumour, not normal brain, and capture / short-read has systematic paralog cross-mapping for homologous families (FGFR1–4, NTRK1–3); a large cohort reduces random error but not this systematic bias.

UCLA long-read

Long-read has low sequencing depth, so it drops out and undercounts moderate / long / degraded transcripts and can under-call a genuinely expressed gene. It is not ground truth, and the long-read-vs-GTEx logFC is itself cross-platform (GTEx is short-read Illumina).

Platform discordance therefore flags uncertainty; it does not prove a gene is or is not expressed, and one platform does not overrule the other. A single-platform high call is hypothesis-generating; an act-on call requires cross-platform concordance plus orthogonal (IHC / protein / fusion) or functional (ex-vivo) support. For a single-platform gene of clinical relevance, the glioma-lineage-specific Caris percentile and an orthogonal assay are the resolving tests.

Fusions (long-read JAFFAL + Caris)

No reported fusion is a confirmed oncogenic driver. The CDK6:VIPR2 / CDK6:KRIT1 cluster is chr7q amplicon-associated rearrangement that corroborates CDK6 amplification, not an oncogenic fusion. CYP46A1:BRAF is hypothesis-only (2 reads, single-platform, confirm before any RAF/MEK inference). QKI fusions = 6q tumour-suppressor disruption; ATRX:MAGT1 = local chrX rearrangement consistent with ATRX inactivation/ALT; XPO1:PPP4R3B is a passenger. No NTRK / FGFR / FGFR3-TACC3 fusion. Full interpretation and the weighting rubric are in the Genomics tab.

Pathway-level signalling (GSEA)

Cascade components (AKT3, PIK3R1, MAPK, mTOR) sit at high absolute short-read mRNA, but gene-set enrichment against normal brain shows PI3K-AKT, MAPK and JAK-STAT all suppressed (NES −1.4 to −1.7, FDR < 0.001), and long-read does not corroborate the RTK / cascade over-expression. The transcriptome does not support an expression-led RTK or downstream-activation argument; the multi-RTK rationale (Biology axes · Multi-RTK) rests on the ex-vivo functional response and is pre-emptive, not constitutive-signalling-driven.

Transcriptome atlas, kinome & signalling landscape

Family-by-family transcriptome census: every RTK family (I–XX, receptors and ligands), every non-receptor tyrosine-kinase family, and the major signalling pathways, with TPM · Rank · Z · Caris %ile · CPM · Rank · Z · logFC vs GTEx · mProbe · Ignite RPPA per gene. Population-normalised Z (Caris log-mean 1.07 / SD 0.59; long-read 0.73 / SD 0.60). An mRNA-abundance census, elevated transcript is not proof of pathway activity.

Analysis of the full census

Highest short-read (Caris Z)

AKT3 +4.71 · PI3K / AKT / mTOR
PIK3R1 +4.67 · PI3K / AKT / mTOR
APC +4.33 · WNT / beta-catenin
NTRK2 +4.26 · VII. TRK / NTRK
PTPN11 +4.17 · RAS / MAPK cascade
TSC2 +4.07 · PI3K / AKT / mTOR
SMAD2 +3.92 · TGF-beta / SMAD / BMP
PTEN +3.88 · PI3K / AKT / mTOR
PDGFRA +3.87 · III. PDGFR
JAK1 +3.84 · JAK
CTNNB1 +3.82 · WNT / beta-catenin
CCND2 +3.76 · Cell-cycle / CDK
TP53 +3.74 · p53 / MDM / apoptosis
PIK3R2 +3.68 · PI3K / AKT / mTOR

Highest long-read (Z)

RPS6 +3.99 · PI3K / AKT / mTOR
PRDX1 +3.16 · NRF2 / redox
TXN +3.15 · NRF2 / redox
CDK6 +3.08 · Cell-cycle / CDK
GPX4 +2.94 · NRF2 / redox
CCND2 +2.48 · Cell-cycle / CDK
RHEB +2.42 · PI3K / AKT / mTOR
NTRK2 +2.27 · VII. TRK / NTRK
NOTCH2NLA +2.14 · Notch
GRB2 +2.09 · RAS / MAPK cascade
BAD +2.07 · p53 / MDM / apoptosis
NFKBIA +2.02 · NF-kB
CDKN2A +2.00 · Cell-cycle / CDK
CDKN2C +1.86 · Cell-cycle / CDK

Families with ≥3 members Z>+3

PI3K / AKT / mTOR — PIK3R1, PIK3R2, PTEN, AKT1, AKT2, AKT3, TSC1, TSC2
RAS / MAPK cascade — BRAF, RAF1, MAPK1, NF1, PTPN11
Cell-cycle / CDK — CDK4, CDK6, CCND2, CDKN1B
III. PDGFR — PDGFRA, PDGFRB, CSF1R
V. FGFR — FGFR1, FGFR2, FGFR3
WNT / beta-catenin — CTNNB1, APC, TCF7L2
TGF-beta / SMAD / BMP — SMAD2, SMAD4, BMPR1A
NRF2 / redox — NFE2L2, TXN, PRDX1

Receptor + ligand co-elevation (autocrine, Caris Z>+2): none at this threshold.

mRNA-high / activation-low discordance (Caris Z>+3, Ignite RPPA ≤15%)

EGFR	Caris Z +3.24	RPPA 7%
ERBB3	Caris Z +3.47	RPPA 12%
PDGFRB	Caris Z +3.03	RPPA 7%
FGFR1	Caris Z +3.55	RPPA 10%
FGFR2	Caris Z +3.62	RPPA 10%
FGFR3	Caris Z +3.52	RPPA 10%
NTRK3	Caris Z +3.29	RPPA 13%
AKT1	Caris Z +3.08	RPPA 14%
AKT2	Caris Z +3.34	RPPA 14%
AKT3	Caris Z +4.71	RPPA 14%

Where measured, RPPA activation overrides transcript abundance for target calls.

Reading

The dominant short-read elevations are the cAMP/PKA module (GNAS, PRKAR1A, Axis 09), AKT3 / PIK3R1 (Axis 05, but Ignite activation low), the WNT effector set (APC, CTNNB1, TCF7L2), PTPN11/JAK1 (Axis 05/07), and the CDK6/cyclin-D programme (Axis 01). Long-read is dominated by ribosomal/redox/structural transcripts (RPS6, TXN, PRDX1, CALM, CST3) and the lineage programme (NOTCH2NLA, ASCL1, Axis 06). Receptor-and-ligand co-elevation is limited; the FGFR family shows multiple receptors high with low cognate-ligand expression (consistent with a receptor-driven, not autocrine-ligand-driven, signal). The recurrent abundance-vs-activation gap, high RTK/effector mRNA with low Ignite phospho, is the central interpretive caveat and reproduces across families.

Receptor tyrosine kinase kinome · click a family to expand

I. ERBB / HER15 genes · max Caris Z +3.47 · max LR Z -0.05

		Caris short-read (2024 tissue)				UCLA long-read (2024 tissue)			vs GTEx	mProbe '24	Ignite RPPA
Gene	type	TPM	Rk	Z	%ile	CPM	Rk	Z	logFC	amol/μg	%
EGFR	receptor	996	112	+3.24	46	4.08	8,777	-0.05	+0.07	299	7%
ERBB2	receptor	254	495	+2.24	46	—	—	—	—	ND	2%
ERBB3	receptor	1,368	77	+3.47	82	2.17	11,294	-0.39	-1.29	—	12%
ERBB4	receptor	512	231	+2.75	—	—	—	—	—	ND	—
EGF	ligand	5.15	12,309	-0.48	—	—	—	—	—	—	—
TGFA	ligand	2.63	15,071	-0.86	—	—	—	—	—	—	—
AREG	ligand	2.03	15,948	-1.00	—	0.38	16,015	-0.98	+0.47	—	—
EREG	ligand	0.65	18,387	-1.44	—	—	—	—	—	—	—
EPGN	ligand	0.62	18,435	-1.46	—	—	—	—	—	—	—
BTC	ligand	0.59	18,475	-1.47	—	0.13	18,010	-1.13	-0.51	—	—
HBEGF	ligand	5.41	12,082	-0.45	—	1.78	12,080	-0.48	-4.72	—	—
NRG1	ligand	15	7,274	+0.22	—	2.55	10,512	-0.30	-0.23	—	—
NRG2	ligand	7.80	10,395	-0.22	—	—	—	—	—	—	—
NRG3	ligand	17	6,847	+0.28	—	0.25	17,639	-1.05	+0.33	—	—
NRG4	ligand	3.43	14,031	-0.72	—	—	—	—	—	—	—

II. Insulin / IGF6 genes · max Caris Z +2.56 · max LR Z +0.11

		Caris short-read (2024 tissue)				UCLA long-read (2024 tissue)			vs GTEx	mProbe '24	Ignite RPPA
Gene	type	TPM	Rk	Z	%ile	CPM	Rk	Z	logFC	amol/μg	%
INSR	receptor	33	3,973	+0.78	—	0.25	16,760	-1.05	-4.78	—	—
IGF1R	receptor	393	298	+2.56	—	0.76	15,011	-0.81	-0.61	—	—
INSRR	receptor	2.72	14,931	-0.85	—	—	—	—	—	—	—
INS	ligand	9.15	9,604	-0.11	—	—	—	—	—	—	—
IGF1	ligand	15	7,207	+0.23	—	5.35	7,681	+0.11	+1.62	—	—
IGF2	ligand	47	2,904	+1.02	—	1.27	13,481	-0.62	+0.44	—	—

III. PDGFR12 genes · max Caris Z +3.87 · max LR Z +0.80

		Caris short-read (2024 tissue)				UCLA long-read (2024 tissue)			vs GTEx	mProbe '24	Ignite RPPA
Gene	type	TPM	Rk	Z	%ile	CPM	Rk	Z	logFC	amol/μg	%
PDGFRA	receptor	2,369	35	+3.87	80	15	4,027	+0.80	-1.72	—	—
PDGFRB	receptor	753	155	+3.03	—	3.19	9,748	-0.19	-1.37	—	7%
CSF1R	receptor	1,317	81	+3.44	—	3.70	9,225	-0.10	-1.13	ND	—
KIT	receptor	274	457	+2.30	—	1.91	11,685	-0.45	-2.07	—	—
FLT3	receptor	30	4,417	+0.70	—	—	—	—	—	—	—
PDGFA	ligand	23	5,390	+0.52	—	8.29	6,034	+0.39	-2.39	—	—
PDGFB	ligand	19	6,257	+0.38	—	0.76	15,053	-0.81	-2.46	—	—
PDGFC	ligand	6.34	11,331	-0.35	—	0.89	14,486	-0.76	+0.68	—	—
PDGFD	ligand	6.11	11,488	-0.37	—	1.02	14,190	-0.71	+1.01	—	—
CSF1	ligand	6.87	10,953	-0.30	—	0.13	19,499	-1.13	-1.66	—	—
KITLG	ligand	4.67	12,747	-0.54	—	5.86	7,321	+0.17	+0.79	—	—
FLT3LG	ligand	12	8,243	+0.07	—	0.64	15,515	-0.86	-1.16	—	—

IV. VEGFR8 genes · max Caris Z +2.53 · max LR Z -0.28

		Caris short-read (2024 tissue)				UCLA long-read (2024 tissue)			vs GTEx	mProbe '24	Ignite RPPA
Gene	type	TPM	Rk	Z	%ile	CPM	Rk	Z	logFC	amol/μg	%
FLT1	receptor	379	317	+2.53	—	—	—	—	—	—	—
KDR	receptor	99	1,358	+1.55	—	1.27	13,285	-0.62	-4.15	—	—
FLT4	receptor	32	4,179	+0.74	—	—	—	—	—	—	—
VEGFA	ligand	124	1,078	+1.72	—	0.25	16,533	-1.05	-1.36	—	—
VEGFB	ligand	6.79	11,018	-0.31	—	2.68	10,462	-0.28	-3.25	—	—
VEGFC	ligand	1.76	16,437	-1.07	—	—	—	—	—	—	—
VEGFD	ligand	4.29	13,144	-0.59	—	—	—	—	—	—	—
PGF	ligand	2.49	15,279	-0.89	—	2.17	11,305	-0.39	-0.00	—	—

V. FGFR21 genes · max Caris Z +3.62 · max LR Z +0.71

		Caris short-read (2024 tissue)				UCLA long-read (2024 tissue)			vs GTEx	mProbe '24	Ignite RPPA
Gene	type	TPM	Rk	Z	%ile	CPM	Rk	Z	logFC	amol/μg	%
FGFR1	receptor	1,522	67	+3.55	18	1.27	13,117	-0.62	-1.83	ND	10%
FGFR2	receptor	1,689	59	+3.62	96	3.57	9,292	-0.12	-3.84	ND	10%
FGFR3	receptor	1,463	73	+3.52	90	—	—	—	—	ND	10%
FGFR4	receptor	257	490	+2.25	—	—	—	—	—	ND	10%
FGF1	ligand	149	882	+1.85	—	14	4,374	+0.71	+0.59	—	—
FGF2	ligand	2.92	14,667	-0.81	—	3.57	9,393	-0.12	-0.55	—	—
FGF3	ligand	3.18	14,328	-0.76	—	—	—	—	—	—	—
FGF4	ligand	3.52	13,913	-0.70	—	—	—	—	—	—	—
FGF5	ligand	—	—	—	—	0.38	16,237	-0.98	+0.02	—	—
FGF6	ligand	8.28	10,079	-0.18	—	—	—	—	—	—	—
FGF7	ligand	4.71	12,695	-0.53	—	1.66	12,353	-0.51	-1.29	—	—
FGF8	ligand	2.71	14,951	-0.85	—	—	—	—	—	—	—
FGF9	ligand	8.87	9,756	-0.13	—	4.08	8,792	-0.05	-1.65	—	—
FGF10	ligand	2.13	15,788	-0.97	—	0.25	17,510	-1.05	+0.33	—	—
FGF16	ligand	0.49	18,627	-1.52	—	0.13	19,014	-1.13	+0.17	—	—
FGF17	ligand	3.01	14,542	-0.79	—	0.13	19,044	-1.13	-1.32	—	—
FGF18	ligand	0.60	18,460	-1.46	—	0.13	18,974	-1.13	-0.39	—	—
FGF19	ligand	2.87	14,721	-0.82	—	—	—	—	—	—	—
FGF20	ligand	2.19	15,719	-0.96	—	0.76	15,116	-0.81	+0.47	—	—
FGF22	ligand	4.68	12,737	-0.54	—	5.61	7,494	+0.14	+1.67	—	—
FGF23	ligand	4.71	12,701	-0.53	—	—	—	—	—	—	—

VI. PTK7 / CCK41 genes · max Caris Z +0.17 · max LR Z -0.14

		Caris short-read (2024 tissue)				UCLA long-read (2024 tissue)			vs GTEx	mProbe '24	Ignite RPPA
Gene	type	TPM	Rk	Z	%ile	CPM	Rk	Z	logFC	amol/μg	%
PTK7	receptor	14	7,601	+0.17	—	3.44	9,541	-0.14	+1.65	—	—

VII. TRK / NTRK7 genes · max Caris Z +4.26 · max LR Z +2.27

		Caris short-read (2024 tissue)				UCLA long-read (2024 tissue)			vs GTEx	mProbe '24	Ignite RPPA
Gene	type	TPM	Rk	Z	%ile	CPM	Rk	Z	logFC	amol/μg	%
NTRK1	receptor	220	570	+2.14	96	—	—	—	—	ND	39%
NTRK2	receptor	4,060	16	+4.26	76	127	621	+2.27	+2.23	—	39%
NTRK3	receptor	1,074	98	+3.29	54	3.31	9,631	-0.16	-3.45	—	13%
NGF	ligand	2.15	15,762	-0.97	—	0.25	17,307	-1.05	-0.29	—	—
BDNF	ligand	0.61	18,452	-1.46	—	3.57	9,411	-0.12	+0.17	—	—
NTF3	ligand	1.85	16,249	-1.04	—	0.13	19,311	-1.13	-0.04	—	—
NTF4	ligand	7.67	10,464	-0.23	—	—	—	—	—	—	—

VIII. ROR4 genes · max Caris Z -0.30 · max LR Z -0.92

		Caris short-read (2024 tissue)				UCLA long-read (2024 tissue)			vs GTEx	mProbe '24	Ignite RPPA
Gene	type	TPM	Rk	Z	%ile	CPM	Rk	Z	logFC	amol/μg	%
ROR1	receptor	1.24	17,379	-1.22	—	0.25	17,596	-1.05	+0.14	—	—
ROR2	receptor	1.58	16,771	-1.12	—	—	—	—	—	—	—
WNT5A	ligand	3.56	13,873	-0.70	—	0.25	17,061	-1.05	-1.01	—	—
WNT5B	ligand	6.92	10,913	-0.30	—	0.51	15,615	-0.92	+0.15	—	—

IX. MuSK2 genes · max Caris Z +0.41 · max LR Z -0.48

		Caris short-read (2024 tissue)				UCLA long-read (2024 tissue)			vs GTEx	mProbe '24	Ignite RPPA
Gene	type	TPM	Rk	Z	%ile	CPM	Rk	Z	logFC	amol/μg	%
MUSK	receptor	2.44	15,357	-0.90	—	—	—	—	—	—	—
AGRN	ligand	20	6,038	+0.41	—	1.78	11,940	-0.48	-3.37	—	—

X. MET / RON4 genes · max Caris Z +1.93 · max LR Z -1.05

		Caris short-read (2024 tissue)				UCLA long-read (2024 tissue)			vs GTEx	mProbe '24	Ignite RPPA
Gene	type	TPM	Rk	Z	%ile	CPM	Rk	Z	logFC	amol/μg	%
MET	receptor	165	792	+1.93	64	—	—	—	—	—	20%
MST1R	receptor	44	3,080	+0.97	—	—	—	—	—	—	—
HGF	ligand	13	7,859	+0.13	—	0.25	16,743	-1.05	-0.55	ND	—
MST1	ligand	27	4,750	+0.63	—	—	—	—	—	—	—

XI. AXL / TAM5 genes · max Caris Z +1.77 · max LR Z +0.74

		Caris short-read (2024 tissue)				UCLA long-read (2024 tissue)			vs GTEx	mProbe '24	Ignite RPPA
Gene	type	TPM	Rk	Z	%ile	CPM	Rk	Z	logFC	amol/μg	%
AXL	receptor	46	2,990	+1.00	—	1.78	11,845	-0.48	-1.67	—	—
TYRO3	receptor	132	993	+1.77	—	14	4,235	+0.74	-0.96	—	—
MERTK	receptor	75	1,790	+1.36	—	10	5,220	+0.54	+3.10	—	—
GAS6	ligand	13	8,025	+0.11	—	12	4,671	+0.66	+1.36	—	—
PROS1	ligand	3.17	14,355	-0.76	—	1.15	13,569	-0.67	+0.57	—	—

XII. TIE5 genes · max Caris Z +0.05 · max LR Z -0.28

		Caris short-read (2024 tissue)				UCLA long-read (2024 tissue)			vs GTEx	mProbe '24	Ignite RPPA
Gene	type	TPM	Rk	Z	%ile	CPM	Rk	Z	logFC	amol/μg	%
TIE1	receptor	12	8,417	+0.05	—	0.13	19,369	-1.13	-2.27	—	—
TEK	receptor	12	8,439	+0.05	—	0.38	16,210	-0.98	-2.83	—	—
ANGPT1	ligand	6.61	11,146	-0.32	—	2.68	10,399	-0.28	+1.88	—	—
ANGPT2	ligand	6.09	11,512	-0.38	—	0.13	18,931	-1.13	-1.76	—	—
ANGPT4	ligand	1.80	16,350	-1.05	—	—	—	—	—	—	—

XIII. EPH22 genes · max Caris Z +2.04 · max LR Z +0.56

		Caris short-read (2024 tissue)				UCLA long-read (2024 tissue)			vs GTEx	mProbe '24	Ignite RPPA
Gene	type	TPM	Rk	Z	%ile	CPM	Rk	Z	logFC	amol/μg	%
EPHA1	receptor	12	8,253	+0.07	—	—	—	—	—	—	—
EPHA2	receptor	192	676	+2.04	80	—	—	—	—	ND	—
EPHA3	receptor	17	6,616	+0.32	—	1.27	13,330	-0.62	-0.61	—	—
EPHA4	receptor	33	4,005	+0.77	—	3.06	9,953	-0.21	-1.86	—	—
EPHA5	receptor	19	6,165	+0.39	—	—	—	—	—	—	—
EPHA6	receptor	12	8,447	+0.05	—	3.70	9,255	-0.10	+1.49	—	—
EPHA7	receptor	16	7,117	+0.24	—	6.12	7,173	+0.20	-1.92	—	—
EPHA8	receptor	2.87	14,728	-0.82	—	—	—	—	—	—	—
EPHA10	receptor	5.95	11,608	-0.39	—	0.25	16,296	-1.05	-2.75	—	—
EPHB1	receptor	37	3,663	+0.84	—	4.33	8,613	-0.01	-2.90	—	—
EPHB2	receptor	16	6,849	+0.28	—	—	—	—	—	—	—
EPHB3	receptor	35	3,798	+0.81	—	—	—	—	—	—	—
EPHB4	receptor	41	3,318	+0.92	—	0.76	15,063	-0.81	+0.10	—	—
EPHB6	receptor	17	6,644	+0.31	—	4.59	8,282	+0.02	-1.10	—	—
EFNA1	ligand	6.93	10,905	-0.29	—	10	5,306	+0.53	-0.59	—	—
EFNA2	ligand	0.98	17,771	-1.31	—	—	—	—	—	—	—
EFNA3	ligand	0.39	18,830	-1.57	—	11	5,137	+0.56	-0.51	—	—
EFNA4	ligand	2.33	15,499	-0.93	—	2.04	11,511	-0.42	+1.04	—	—
EFNA5	ligand	6.91	10,921	-0.30	—	3.31	9,656	-0.16	+1.54	—	—
EFNB1	ligand	2.71	14,957	-0.85	—	2.80	10,325	-0.25	+0.07	—	—
EFNB2	ligand	2.74	14,911	-0.84	—	2.04	11,518	-0.42	+0.12	—	—
EFNB3	ligand	3.51	13,928	-0.71	—	1.02	14,266	-0.71	-4.06	—	—

XIV. RET9 genes · max Caris Z +1.96 · max LR Z +0.40

		Caris short-read (2024 tissue)				UCLA long-read (2024 tissue)			vs GTEx	mProbe '24	Ignite RPPA
Gene	type	TPM	Rk	Z	%ile	CPM	Rk	Z	logFC	amol/μg	%
RET	receptor	171	762	+1.96	88	0.25	16,543	-1.05	-0.96	ND	9%
GDNF	ligand	0.85	18,014	-1.36	—	—	—	—	—	—	—
NRTN	ligand	1.34	17,190	-1.19	—	—	—	—	—	—	—
ARTN	ligand	4.07	13,361	-0.62	—	—	—	—	—	—	—
PSPN	ligand	4.62	12,787	-0.55	—	8.41	5,972	+0.40	-0.15	—	—
GFRA1	ligand	2.23	15,660	-0.95	—	0.13	19,902	-1.13	-0.75	—	—
GFRA2	ligand	9.24	9,559	-0.11	—	4.08	8,813	-0.05	+0.93	—	—
GFRA3	ligand	1.41	17,044	-1.16	—	—	—	—	—	—	—
GFRA4	ligand	1.31	17,237	-1.20	—	—	—	—	—	—	—

XV. RYK4 genes · max Caris Z +0.34 · max LR Z +0.22

		Caris short-read (2024 tissue)				UCLA long-read (2024 tissue)			vs GTEx	mProbe '24	Ignite RPPA
Gene	type	TPM	Rk	Z	%ile	CPM	Rk	Z	logFC	amol/μg	%
RYK	receptor	18	6,502	+0.34	—	1.78	12,006	-0.48	-0.96	—	—
WNT1	ligand	1.14	17,539	-1.25	—	—	—	—	—	—	—
WNT3	ligand	2.01	15,988	-1.00	—	6.37	7,030	+0.22	+0.07	—	—
WNT3A	ligand	1.30	17,246	-1.20	—	—	—	—	—	—	—

XVI. DDR6 genes · max Caris Z +2.81 · max LR Z +0.38

		Caris short-read (2024 tissue)				UCLA long-read (2024 tissue)			vs GTEx	mProbe '24	Ignite RPPA
Gene	type	TPM	Rk	Z	%ile	CPM	Rk	Z	logFC	amol/μg	%
DDR1	receptor	504	236	+2.74	—	8.16	6,126	+0.38	+0.53	—	—
DDR2	receptor	554	214	+2.81	—	6.25	7,143	+0.21	-0.05	—	—
COL1A1	ligand	23	5,362	+0.52	—	—	—	—	—	—	—
COL1A2	ligand	10	8,960	-0.03	—	—	—	—	—	—	—
COL3A1	ligand	4.37	13,068	-0.58	—	0.51	15,876	-0.92	+0.59	—	—
COL5A1	ligand	10	8,955	-0.03	—	—	—	—	—	—	—

XVII. LMR / AATYK3 genes · max Caris Z +0.63 · max LR Z -0.25

		Caris short-read (2024 tissue)				UCLA long-read (2024 tissue)			vs GTEx	mProbe '24	Ignite RPPA
Gene	type	TPM	Rk	Z	%ile	CPM	Rk	Z	logFC	amol/μg	%
AATK	receptor	21	5,754	+0.46	—	2.80	10,302	-0.25	+0.06	—	—
LMTK2	receptor	7.39	10,621	-0.25	—	1.02	14,067	-0.71	-5.46	—	—
LMTK3	receptor	27	4,758	+0.63	—	—	—	—	—	—	—

XVIII. STYK11 genes · max Caris Z -1.04 · max LR Z -0.55

		Caris short-read (2024 tissue)				UCLA long-read (2024 tissue)			vs GTEx	mProbe '24	Ignite RPPA
Gene	type	TPM	Rk	Z	%ile	CPM	Rk	Z	logFC	amol/μg	%
STYK1	receptor	1.84	16,279	-1.04	—	1.53	12,521	-0.55	-1.36	—	—

XIX. ALK / LTK4 genes · max Caris Z +1.72 · max LR Z -1.13

		Caris short-read (2024 tissue)				UCLA long-read (2024 tissue)			vs GTEx	mProbe '24	Ignite RPPA
Gene	type	TPM	Rk	Z	%ile	CPM	Rk	Z	logFC	amol/μg	%
ALK	receptor	123	1,083	+1.72	68	—	—	—	—	ND	13%
LTK	receptor	15	7,297	+0.21	—	—	—	—	—	—	—
ALKAL1	ligand	2.78	14,851	-0.84	—	0.13	19,270	-1.13	+0.17	—	—
ALKAL2	ligand	7.83	10,370	-0.22	—	—	—	—	—	—	—

XX. ROS1 genes · max Caris Z +0.95 · max LR Z —

		Caris short-read (2024 tissue)				UCLA long-read (2024 tissue)			vs GTEx	mProbe '24	Ignite RPPA
Gene	type	TPM	Rk	Z	%ile	CPM	Rk	Z	logFC	amol/μg	%
ROS1	receptor	42	3,203	+0.95	92	—	—	—	—	ND	46%

Non-receptor tyrosine kinases · click a family to expand

SRC family9 genes · max Caris Z +3.06 · max LR Z +0.43

		Caris short-read (2024 tissue)				UCLA long-read (2024 tissue)			vs GTEx	mProbe '24	Ignite RPPA
Gene	type	TPM	Rk	Z	%ile	CPM	Rk	Z	logFC	amol/μg	%
SRC	kinase	461	255	+2.67	90	2.42	10,762	-0.33	+0.42	306	6%
YES1	kinase	49	2,800	+1.04	—	8.92	5,811	+0.43	+2.59	—	—
FYN	kinase	777	148	+3.06	—	3.70	9,176	-0.10	-0.84	—	—
LYN	kinase	79	1,695	+1.40	—	1.91	11,681	-0.45	-0.45	—	—
LCK	kinase	18	6,407	+0.35	—	—	—	—	—	—	—
HCK	kinase	8.34	10,035	-0.17	—	0.25	16,909	-1.05	-0.41	—	—
FGR	kinase	9.93	9,182	-0.06	—	—	—	—	—	—	—
BLK	kinase	4.40	13,033	-0.58	—	—	—	—	—	—	—
FRK	kinase	0.65	18,380	-1.44	—	—	—	—	—	—	—

ABL2 genes · max Caris Z +2.22 · max LR Z -1.05

		Caris short-read (2024 tissue)				UCLA long-read (2024 tissue)			vs GTEx	mProbe '24	Ignite RPPA
Gene	type	TPM	Rk	Z	%ile	CPM	Rk	Z	logFC	amol/μg	%
ABL1	kinase	248	515	+2.22	—	0.25	16,892	-1.05	-2.36	—	—
ABL2	kinase	70	1,929	+1.31	—	—	—	—	—	—	—

JAK4 genes · max Caris Z +3.84 · max LR Z -0.05

		Caris short-read (2024 tissue)				UCLA long-read (2024 tissue)			vs GTEx	mProbe '24	Ignite RPPA
Gene	type	TPM	Rk	Z	%ile	CPM	Rk	Z	logFC	amol/μg	%
JAK1	kinase	2,273	37	+3.84	—	4.08	8,789	-0.05	-2.88	146	—
JAK2	kinase	200	635	+2.07	—	2.29	11,108	-0.36	-0.71	—	5%
JAK3	kinase	145	900	+1.84	—	—	—	—	—	—	—
TYK2	kinase	150	877	+1.86	—	3.82	9,114	-0.08	+0.01	—	—

FAK2 genes · max Caris Z +2.85 · max LR Z +0.67

		Caris short-read (2024 tissue)				UCLA long-read (2024 tissue)			vs GTEx	mProbe '24	Ignite RPPA
Gene	type	TPM	Rk	Z	%ile	CPM	Rk	Z	logFC	amol/μg	%
PTK2	kinase	584	205	+2.85	—	13	4,582	+0.67	-0.93	—	—
PTK2B	kinase	203	623	+2.08	—	3.70	9,218	-0.10	-0.97	—	—

FES2 genes · max Caris Z +0.92 · max LR Z -1.13

		Caris short-read (2024 tissue)				UCLA long-read (2024 tissue)			vs GTEx	mProbe '24	Ignite RPPA
Gene	type	TPM	Rk	Z	%ile	CPM	Rk	Z	logFC	amol/μg	%
FES	kinase	16	7,037	+0.25	—	0.13	18,200	-1.13	-2.29	—	—
FER	kinase	41	3,357	+0.92	—	—	—	—	—	—	—

TEC5 genes · max Caris Z +1.69 · max LR Z -1.13

		Caris short-read (2024 tissue)				UCLA long-read (2024 tissue)			vs GTEx	mProbe '24	Ignite RPPA
Gene	type	TPM	Rk	Z	%ile	CPM	Rk	Z	logFC	amol/μg	%
BTK	kinase	119	1,121	+1.69	—	0.13	17,775	-1.13	-0.21	—	—
BMX	kinase	1.10	17,593	-1.26	—	—	—	—	—	—	—
ITK	kinase	12	8,364	+0.06	—	—	—	—	—	—	—
TEC	kinase	2.14	15,785	-0.97	—	—	—	—	—	—	—
TXK	kinase	4.56	12,867	-0.55	—	—	—	—	—	—	—

SYK2 genes · max Caris Z +0.49 · max LR Z -1.13

		Caris short-read (2024 tissue)				UCLA long-read (2024 tissue)			vs GTEx	mProbe '24	Ignite RPPA
Gene	type	TPM	Rk	Z	%ile	CPM	Rk	Z	logFC	amol/μg	%
SYK	kinase	22	5,553	+0.49	—	0.13	19,672	-1.13	-0.20	—	—
ZAP70	kinase	7.24	10,712	-0.27	—	—	—	—	—	—	—

CSK2 genes · max Caris Z +1.00 · max LR Z -0.36

		Caris short-read (2024 tissue)				UCLA long-read (2024 tissue)			vs GTEx	mProbe '24	Ignite RPPA
Gene	type	TPM	Rk	Z	%ile	CPM	Rk	Z	logFC	amol/μg	%
CSK	kinase	46	2,984	+1.00	—	1.78	11,969	-0.48	-0.39	—	—
MATK	kinase	24	5,304	+0.53	—	2.29	10,939	-0.36	-3.21	—	—

ACK2 genes · max Caris Z +1.60 · max LR Z -0.67

		Caris short-read (2024 tissue)				UCLA long-read (2024 tissue)			vs GTEx	mProbe '24	Ignite RPPA
Gene	type	TPM	Rk	Z	%ile	CPM	Rk	Z	logFC	amol/μg	%
TNK1	kinase	8.96	9,708	-0.13	—	—	—	—	—	—	—
TNK2	kinase	105	1,279	+1.60	—	1.15	13,683	-0.67	-4.18	—	—

BRK3 genes · max Caris Z -0.57 · max LR Z —

		Caris short-read (2024 tissue)				UCLA long-read (2024 tissue)			vs GTEx	mProbe '24	Ignite RPPA
Gene	type	TPM	Rk	Z	%ile	CPM	Rk	Z	logFC	amol/μg	%
PTK6	kinase	4.43	13,000	-0.57	—	—	—	—	—	—	—
FRK	kinase	0.65	18,380	-1.44	—	—	—	—	—	—	—
SRMS	kinase	2.27	15,596	-0.94	—	—	—	—	—	—	—

Signalling pathways · click a pathway to expand

RAS / MAPK cascade26 genes · max Caris Z +4.17 · max LR Z +2.09

		Caris short-read (2024 tissue)				UCLA long-read (2024 tissue)			vs GTEx	mProbe '24	Ignite RPPA
Gene	type	TPM	Rk	Z	%ile	CPM	Rk	Z	logFC	amol/μg	%
KRAS		270	466	+2.28	78	50	1,524	+1.61	+0.98	2,290	—
NRAS		163	806	+1.92	24	12	4,845	+0.61	+0.21	557	—
HRAS		581	207	+2.84	95	12	4,819	+0.62	-0.89	—	—
RIT1		151	868	+1.86	—	26	2,676	+1.16	+0.92	—	—
RRAS		3.94	13,491	-0.64	—	1.53	12,701	-0.55	-2.87	—	—
RRAS2		60	2,252	+1.19	—	0.25	16,501	-1.05	-1.15	—	—
MRAS		118	1,131	+1.68	—	21	3,258	+0.99	-0.97	—	—
BRAF		923	118	+3.18	80	8.80	5,852	+0.43	-1.18	87	52%
RAF1		870	131	+3.14	—	11	5,130	+0.56	-1.05	—	—
ARAF		323	374	+2.42	—	12	4,731	+0.64	+0.29	—	—
MAP2K1		693	169	+2.97	—	5.35	7,723	+0.11	-6.84	—	12%
MAP2K2		714	167	+2.99	—	10	5,249	+0.54	-0.90	—	12%
MAPK1		1,531	66	+3.55	—	6.12	7,242	+0.20	-5.54	—	89%
MAPK3		235	537	+2.18	—	9.56	5,555	+0.48	-1.30	—	89%
MAPK7		15	7,322	+0.21	—	5.23	7,808	+0.10	+2.11	—	—
NF1		920	120	+3.18	—	18	3,542	+0.91	+1.19	—	—
SHC1		11	8,742	+0.00	—	12	4,846	+0.61	+2.10	—	—
GRB2		135	962	+1.78	—	99	781	+2.09	+1.17	—	—
PTPN11		3,557	21	+4.17	—	6.12	7,229	+0.20	-2.80	—	—
SPRY1		5.04	12,435	-0.49	—	6.50	6,945	+0.23	+0.15	—	—
SPRY2		11	8,668	+0.01	—	12	4,683	+0.65	-0.12	—	—
SPRY4		1.01	17,729	-1.30	—	1.53	12,578	-0.55	-0.62	—	—
SPRED1		122	1,098	+1.71	—	3.06	9,903	-0.21	-2.61	—	—
DUSP4		—	—	—	—	1.02	13,956	-0.71	-1.95	—	—
DUSP6		—	—	—	—	13	4,527	+0.68	-0.07	—	—
RASA1		120	1,113	+1.70	—	0.38	16,295	-0.98	-1.53	—	—

PI3K / AKT / mTOR24 genes · max Caris Z +4.71 · max LR Z +3.99

		Caris short-read (2024 tissue)				UCLA long-read (2024 tissue)			vs GTEx	mProbe '24	Ignite RPPA
Gene	type	TPM	Rk	Z	%ile	CPM	Rk	Z	logFC	amol/μg	%
PIK3CA		403	289	+2.58	64	0.51	15,744	-0.92	-2.28	635	—
PIK3CB		725	165	+3.00	—	1.66	12,201	-0.51	+0.65	254	—
PIK3CD		49	2,762	+1.05	—	—	—	—	—	—	—
PIK3CG		12	8,323	+0.06	—	—	—	—	—	—	—
PIK3R1		7,095	5.00	+4.67	—	13	4,525	+0.68	-0.85	—	—
PIK3R2		1,820	51	+3.68	—	0.13	19,662	-1.13	-4.74	—	—
PIK3R3		3.19	14,316	-0.76	—	2.04	11,365	-0.42	-2.06	—	—
PTEN		2,415	34	+3.88	80	2.80	10,163	-0.25	-2.95	—	—
AKT1		803	143	+3.08	—	0.25	16,622	-1.05	-0.65	—	14%
AKT2		1,155	89	+3.34	—	33	2,201	+1.32	+0.42	—	14%
AKT3		7,501	3.00	+4.71	—	1.15	13,649	-0.67	-4.22	—	14%
MTOR		500	238	+2.73	58	3.57	9,304	-0.12	+0.17	—	22%
RPTOR		66	2,036	+1.26	—	1.40	12,885	-0.59	-1.42	—	—
RICTOR		123	1,091	+1.71	—	2.80	10,185	-0.25	-1.67	—	—
TSC1		837	135	+3.11	78	12	4,797	+0.63	+1.80	—	—
TSC2		3,103	26	+4.07	75	7.39	6,479	+0.31	+0.23	—	—
RHEB		349	342	+2.47	—	157	496	+2.42	-0.29	—	—
PDPK1		23	5,367	+0.52	—	10	5,281	+0.53	-1.73	—	—
RPS6KB1		62	2,170	+1.22	—	2.17	11,226	-0.39	-1.45	—	—
RPS6		390	302	+2.55	—	1,391	28	+3.99	-0.91	—	—
EIF4EBP1		—	—	—	—	8.67	5,866	+0.42	+1.21	—	—
INPP4B		70	1,926	+1.31	—	1.78	12,103	-0.48	-1.25	—	—
PHLPP1		—	—	—	—	4.33	8,575	-0.01	-1.38	—	—
PIK3C2B		102	1,315	+1.58	—	7.14	6,598	+0.29	+2.61	—	—

JAK / STAT18 genes · max Caris Z +3.84 · max LR Z +1.20

		Caris short-read (2024 tissue)				UCLA long-read (2024 tissue)			vs GTEx	mProbe '24	Ignite RPPA
Gene	type	TPM	Rk	Z	%ile	CPM	Rk	Z	logFC	amol/μg	%
JAK1		2,273	37	+3.84	—	4.08	8,789	-0.05	-2.88	146	—
JAK2		200	635	+2.07	—	2.29	11,108	-0.36	-0.71	—	5%
JAK3		145	900	+1.84	—	—	—	—	—	—	—
TYK2		150	877	+1.86	—	3.82	9,114	-0.08	+0.01	—	—
STAT1		28	4,645	+0.65	—	7.27	6,555	+0.30	+1.34	1,222	—
STAT2		32	4,144	+0.75	—	3.06	9,933	-0.21	-2.31	—	—
STAT3		887	125	+3.15	—	23	2,951	+1.08	-0.68	784	18%
STAT4		9.45	9,434	-0.09	—	2.29	11,072	-0.36	+0.03	—	—
STAT5A		25	5,051	+0.58	—	—	—	—	—	—	17%
STAT5B		93	1,460	+1.51	—	1.53	12,507	-0.55	-4.66	—	17%
STAT6		41	3,299	+0.93	—	1.27	13,492	-0.62	-2.93	—	—
SOCS1		41	3,314	+0.93	—	0.13	18,855	-1.13	-0.36	—	—
SOCS3		0.72	18,264	-1.41	—	—	—	—	—	—	—
CISH		3.12	14,408	-0.77	—	0.89	14,416	-0.76	+0.42	—	—
PTPN6		8.70	9,834	-0.15	—	0.13	18,287	-1.13	-0.44	—	—
IL6ST		107	1,253	+1.61	—	28	2,564	+1.20	+1.47	—	—
OSMR		7.93	10,295	-0.21	—	1.53	12,573	-0.55	-0.98	—	—
LIFR		498	239	+2.73	—	4.33	8,594	-0.01	-1.53	—	—

WNT / beta-catenin27 genes · max Caris Z +4.33 · max LR Z +1.61

		Caris short-read (2024 tissue)				UCLA long-read (2024 tissue)			vs GTEx	mProbe '24	Ignite RPPA
Gene	type	TPM	Rk	Z	%ile	CPM	Rk	Z	logFC	amol/μg	%
WNT5A		3.56	13,873	-0.70	—	0.25	17,061	-1.05	-1.01	—	—
WNT5B		6.92	10,913	-0.30	—	0.51	15,615	-0.92	+0.15	—	—
WNT7B		4.67	12,751	-0.54	—	—	—	—	—	—	—
CTNNB1		2,209	39	+3.82	—	23	2,920	+1.08	+0.70	—	—
APC		4,462	13	+4.33	—	27	2,623	+1.17	+1.33	—	—
AXIN1		42	3,235	+0.94	—	—	—	—	—	—	—
AXIN2		458	258	+2.67	—	4.33	8,578	-0.01	-0.29	—	—
GSK3B		308	398	+2.38	—	6.25	7,104	+0.21	-0.99	—	—
LRP5		31	4,292	+0.72	—	—	—	—	—	—	—
LRP6		61	2,202	+1.21	—	1.66	12,143	-0.51	-1.41	—	—
FZD1		1.71	16,518	-1.08	—	13	4,450	+0.70	+2.26	—	—
FZD2		1.54	16,830	-1.13	—	0.25	16,799	-1.05	-0.52	—	—
FZD3		2.41	15,400	-0.91	—	12	4,842	+0.61	+0.39	—	—
FZD6		5.76	11,763	-0.41	—	0.25	16,681	-1.05	-0.03	—	—
FZD7		1.51	16,870	-1.13	—	20	3,279	+0.99	+3.13	—	—
TCF7		12	8,299	+0.07	—	0.13	19,028	-1.13	-0.15	—	—
TCF7L1		6.93	10,901	-0.29	—	0.13	19,007	-1.13	-1.80	—	—
TCF7L2		1,585	61	+3.58	—	1.91	11,775	-0.45	-0.33	—	—
LEF1		13	8,068	+0.10	—	0.25	17,353	-1.05	-0.01	—	—
DVL1		13	7,809	+0.14	—	—	—	—	—	—	—
DVL3		89	1,502	+1.48	—	4.59	8,376	+0.02	-2.09	—	—
CSNK1A1		64	2,092	+1.24	—	33	2,232	+1.31	+0.31	—	—
RNF43		90	1,492	+1.49	—	—	—	—	—	—	—
ZNRF3		33	4,012	+0.77	—	0.13	17,722	-1.13	-3.76	—	—
PORCN		6.23	11,404	-0.36	—	6.12	7,203	+0.20	-0.12	—	—
RSPO2		42	3,278	+0.93	—	4.08	8,856	-0.05	+2.10	—	—
SFRP1		—	—	—	—	50	1,518	+1.61	+1.87	—	—

Hippo20 genes · max Caris Z +3.21 · max LR Z +1.02

		Caris short-read (2024 tissue)				UCLA long-read (2024 tissue)			vs GTEx	mProbe '24	Ignite RPPA
Gene	type	TPM	Rk	Z	%ile	CPM	Rk	Z	logFC	amol/μg	%
STK3		40	3,399	+0.90	—	3.06	9,859	-0.21	-0.53	—	—
STK4		34	3,917	+0.79	—	15	4,098	+0.77	-0.33	—	—
SAV1		5.88	11,669	-0.40	—	3.44	9,469	-0.14	-1.45	—	—
LATS1		15	7,316	+0.21	—	3.31	9,565	-0.16	-2.81	—	—
LATS2		8.93	9,726	-0.13	—	3.70	9,151	-0.10	+0.53	—	—
MOB1A		14	7,519	+0.18	—	21	3,145	+1.02	-0.64	—	—
MOB1B		8.66	9,859	-0.15	—	8.41	6,013	+0.40	+0.77	—	—
YAP1		123	1,092	+1.71	—	0.51	15,645	-0.92	-1.54	—	—
WWTR1		214	586	+2.12	—	3.44	9,482	-0.14	-0.71	—	—
TEAD1		24	5,187	+0.55	—	—	—	—	—	—	—
TEAD2		8.97	9,702	-0.13	—	0.13	19,439	-1.13	-0.26	—	—
TEAD3		2.66	15,018	-0.86	—	0.13	19,435	-1.13	-0.57	—	—
TEAD4		3.25	14,257	-0.75	—	—	—	—	—	—	—
NF2		824	136	+3.10	—	2.42	10,818	-0.33	-0.51	—	—
AMOTL2		11	8,754	+0.00	—	3.82	9,111	-0.08	+0.07	—	—
FAT1		956	114	+3.21	—	1.02	14,138	-0.71	-1.65	—	—
FAT4		8.36	10,026	-0.17	—	2.68	10,433	-0.28	+1.54	—	—
DCHS1		5.04	12,431	-0.49	—	1.02	14,074	-0.71	-1.57	—	—
PTPN14		8.36	10,022	-0.17	—	0.13	18,280	-1.13	-3.14	—	—
AJUBA		4.90	12,540	-0.51	—	1.02	14,210	-0.71	+0.68	—	—

NF-kB21 genes · max Caris Z +3.55 · max LR Z +2.02

		Caris short-read (2024 tissue)				UCLA long-read (2024 tissue)			vs GTEx	mProbe '24	Ignite RPPA
Gene	type	TPM	Rk	Z	%ile	CPM	Rk	Z	logFC	amol/μg	%
NFKB1		12	8,365	+0.06	—	—	—	—	—	—	—
NFKB2		34	3,935	+0.78	—	0.25	17,294	-1.05	-0.56	—	—
RELA		89	1,504	+1.48	—	16	3,955	+0.81	+1.84	—	—
RELB		8.78	9,801	-0.14	—	0.13	18,110	-1.13	+0.17	—	—
REL		7.31	10,669	-0.26	—	1.15	13,728	-0.67	-0.56	—	—
NFKBIA		1,532	65	+3.55	—	89	852	+2.02	+1.10	—	—
NFKBIB		20	5,915	+0.43	—	15	4,070	+0.79	-1.28	—	—
NFKBIE		11	8,712	+0.01	—	—	—	—	—	—	—
CHUK		9.38	9,473	-0.10	—	0.89	14,467	-0.76	-0.73	—	—
IKBKB		51	2,688	+1.07	—	2.80	10,151	-0.25	+0.18	—	—
IKBKG		8.38	10,014	-0.17	—	9.82	5,473	+0.50	+0.67	—	—
TNFAIP3		—	—	—	—	0.13	19,630	-1.13	-0.67	—	—
TRAF2		13	8,002	+0.11	—	2.68	10,420	-0.28	-0.88	—	—
TRAF6		6.73	11,061	-0.31	—	7.14	6,626	+0.29	+1.34	—	—
MAP3K7		11	8,698	+0.01	—	8.54	5,963	+0.41	+1.12	—	—
MAP3K14		4.62	12,786	-0.55	—	0.13	19,306	-1.13	-0.75	—	—
BIRC2		19	6,155	+0.39	—	6.63	6,922	+0.25	+0.84	—	—
BIRC3		14	7,546	+0.18	—	—	—	—	—	—	—
TNFRSF1A		33	4,028	+0.77	—	3.19	9,766	-0.19	-0.01	—	—
TBK1		48	2,816	+1.04	—	5.35	7,718	+0.11	-1.10	—	—
RIPK1		41	3,294	+0.93	—	0.25	16,821	-1.05	-1.30	—	—

Notch25 genes · max Caris Z +3.42 · max LR Z +2.14

		Caris short-read (2024 tissue)				UCLA long-read (2024 tissue)			vs GTEx	mProbe '24	Ignite RPPA
Gene	type	TPM	Rk	Z	%ile	CPM	Rk	Z	logFC	amol/μg	%
NOTCH1		1,278	85	+3.42	—	0.13	19,403	-1.13	+0.17	—	—
NOTCH2		196	660	+2.05	—	4.72	8,260	+0.04	-1.11	—	—
NOTCH3		96	1,396	+1.54	—	—	—	—	—	—	—
NOTCH4		11	8,696	+0.01	—	1.02	13,816	-0.71	+0.08	—	—
NOTCH2NLA		75	1,793	+1.35	—	105	732	+2.14	+4.27	—	—
DLL1		18	6,439	+0.35	—	4.97	8,048	+0.07	+0.41	—	—
DLL3		6.94	10,897	-0.29	—	5.10	7,891	+0.08	-2.17	—	—
DLL4		3.51	13,930	-0.71	—	0.51	15,855	-0.92	-3.79	—	—
JAG1		123	1,086	+1.71	—	3.70	9,164	-0.10	-0.48	—	—
JAG2		6.93	10,906	-0.29	—	—	—	—	—	—	—
HES1		18	6,348	+0.36	—	5.23	7,805	+0.10	+0.73	—	—
HES5		6.90	10,937	-0.30	—	2.93	10,085	-0.23	+0.23	—	—
HEY1		92	1,476	+1.50	—	25	2,800	+1.12	+2.03	—	—
HEY2		6.44	11,260	-0.34	—	3.06	9,932	-0.21	+1.28	—	—
RBPJ		192	675	+2.04	—	9.94	5,429	+0.50	-0.18	—	—
NUMB		34	3,929	+0.78	—	10	5,386	+0.51	-0.01	—	—
DTX1		30	4,425	+0.70	—	0.25	17,360	-1.05	-4.11	—	—
MAML1		50	2,702	+1.07	—	1.78	12,010	-0.48	-0.81	—	—
MAML2		88	1,524	+1.47	—	2.80	10,278	-0.25	-0.04	—	—
MAML3		7.28	10,682	-0.26	—	1.27	13,120	-0.62	-2.02	—	—
PSEN1		140	929	+1.81	—	9.43	5,587	+0.47	-1.30	—	—
PSEN2		19	6,291	+0.37	—	7.14	6,628	+0.29	+0.42	—	—
ADAM10		89	1,514	+1.48	—	6.88	6,756	+0.27	-1.13	—	—
ADAM17		50	2,703	+1.07	—	4.59	8,390	+0.02	-0.07	—	—
LFNG		17	6,598	+0.32	—	1.53	12,445	-0.55	-0.41	—	—

Hedgehog15 genes · max Caris Z +3.43 · max LR Z +0.17

		Caris short-read (2024 tissue)				UCLA long-read (2024 tissue)			vs GTEx	mProbe '24	Ignite RPPA
Gene	type	TPM	Rk	Z	%ile	CPM	Rk	Z	logFC	amol/μg	%
SHH		6.33	11,343	-0.35	—	0.13	19,209	-1.13	-0.39	—	—
IHH		1.91	16,150	-1.03	—	—	—	—	—	—	—
DHH		0.74	18,220	-1.40	—	—	—	—	—	—	—
PTCH1		1,300	82	+3.43	—	4.21	8,690	-0.03	-1.48	—	—
PTCH2		13	8,042	+0.10	—	0.13	18,370	-1.13	-2.49	—	—
SMO		250	510	+2.23	—	0.13	17,836	-1.13	-1.05	—	—
GLI1		16	6,986	+0.26	—	—	—	—	—	—	—
GLI2		35	3,793	+0.81	—	1.02	13,970	-0.71	+1.01	—	—
GLI3		23	5,430	+0.51	—	2.29	11,104	-0.36	+0.42	—	—
SUFU		286	432	+2.33	—	1.91	11,625	-0.45	-1.88	—	—
HHIP		36	3,711	+0.83	—	3.31	9,669	-0.16	-0.72	—	—
BOC		54	2,498	+1.12	—	4.46	8,429	+0.01	+1.07	—	—
CDON		5.82	11,714	-0.40	—	0.13	18,421	-1.13	-1.90	—	—
GAS1		2.60	15,118	-0.87	—	5.86	7,330	+0.17	+1.51	—	—
KIF7		11	8,509	+0.04	—	1.78	12,086	-0.48	+0.71	—	—

TGF-beta / SMAD / BMP22 genes · max Caris Z +3.92 · max LR Z +1.49

		Caris short-read (2024 tissue)				UCLA long-read (2024 tissue)			vs GTEx	mProbe '24	Ignite RPPA
Gene	type	TPM	Rk	Z	%ile	CPM	Rk	Z	logFC	amol/μg	%
TGFB1		11	8,551	+0.03	—	1.27	13,100	-0.62	-0.24	—	—
TGFB2		33	4,062	+0.76	—	15	4,144	+0.76	+2.51	—	—
TGFB3		11	8,710	+0.01	—	2.55	10,608	-0.30	-3.27	—	—
TGFBR1		45	3,033	+0.99	—	—	—	—	—	—	—
TGFBR2		42	3,210	+0.95	—	0.13	19,296	-1.13	-0.04	—	—
TGFBR3		17	6,678	+0.31	—	0.76	15,249	-0.81	+0.48	—	—
SMAD2		2,541	31	+3.92	—	14	4,213	+0.74	-0.69	—	—
SMAD3		139	937	+1.80	—	2.04	11,398	-0.42	-2.77	—	—
SMAD4		1,002	111	+3.24	—	2.42	10,739	-0.33	-0.24	—	—
SMAD7		8.26	10,089	-0.18	—	0.25	16,536	-1.05	-2.62	—	—
BMP2		2.84	14,764	-0.82	—	1.02	13,921	-0.71	+0.27	—	—
BMP4		1.66	16,626	-1.09	—	0.76	15,146	-0.81	-0.03	—	—
BMP7		47	2,885	+1.02	—	42	1,788	+1.49	+0.27	—	—
BMPR1A		804	142	+3.08	—	1.15	13,566	-0.67	-2.17	—	—
BMPR2		23	5,439	+0.51	—	8.41	6,022	+0.40	+1.23	—	—
ACVR1		234	544	+2.18	—	1.53	12,571	-0.55	+0.92	—	—
ACVR2A		9.55	9,370	-0.09	—	15	4,011	+0.80	+0.94	—	—
INHBA		17	6,794	+0.29	—	—	—	—	—	—	—
GDF15		6.72	11,066	-0.31	—	1.02	14,224	-0.71	+0.53	—	—
LTBP1		—	—	—	—	1.40	12,995	-0.59	+1.08	—	—
SKI		14	7,642	+0.16	—	0.64	15,385	-0.86	-2.31	—	—
SKIL		35	3,784	+0.82	—	6.37	7,010	+0.22	-0.39	—	—

NRF2 / redox17 genes · max Caris Z +3.10 · max LR Z +3.16

		Caris short-read (2024 tissue)				UCLA long-read (2024 tissue)			vs GTEx	mProbe '24	Ignite RPPA
Gene	type	TPM	Rk	Z	%ile	CPM	Rk	Z	logFC	amol/μg	%
NFE2L2		824	137	+3.10	—	40	1,853	+1.46	+1.90	—	—
KEAP1		—	—	—	—	26	2,698	+1.15	+1.88	—	—
CUL3		180	726	+1.99	—	1.27	13,195	-0.62	-1.37	—	—
NQO1		8.14	10,160	-0.19	—	6.12	7,199	+0.20	-2.97	—	—
GCLC		75	1,798	+1.35	—	2.29	10,968	-0.36	-2.33	—	—
GCLM		3.00	14,568	-0.79	—	9.82	5,448	+0.50	+0.37	—	—
TXN		57	2,362	+1.16	—	433	171	+3.15	-0.58	—	—
TXNRD1		55	2,445	+1.14	—	6.25	7,146	+0.21	+0.90	—	—
SLC7A11		34	3,874	+0.80	—	12	4,701	+0.64	+0.20	—	—
GPX4		275	455	+2.30	—	324	239	+2.94	-1.06	—	—
PRDX1		33	4,063	+0.76	—	441	168	+3.16	+0.02	—	—
SRXN1		2.35	15,470	-0.92	—	0.25	17,207	-1.05	-4.13	—	—
GSR		48	2,853	+1.03	—	5.61	7,540	+0.14	-2.42	—	—
HMOX1		7.75	10,430	-0.22	—	3.82	9,042	-0.08	-0.25	—	—
G6PD		21	5,859	+0.44	—	5.61	7,500	+0.14	+0.65	—	—
ME1		4.54	12,885	-0.56	—	8.54	5,954	+0.41	-0.82	—	—
IDH1		457	259	+2.67	—	33	2,212	+1.32	+2.33	—	—

p53 / MDM / apoptosis21 genes · max Caris Z +3.74 · max LR Z +2.07

		Caris short-read (2024 tissue)				UCLA long-read (2024 tissue)			vs GTEx	mProbe '24	Ignite RPPA
Gene	type	TPM	Rk	Z	%ile	CPM	Rk	Z	logFC	amol/μg	%
TP53		1,979	46	+3.74	76	4.84	8,110	+0.05	-0.03	ND	—
MDM2		225	558	+2.15	—	8.16	6,098	+0.38	-1.19	—	—
MDM4		408	283	+2.59	—	36	2,039	+1.38	+1.56	—	—
BCL2		112	1,196	+1.65	—	12	4,682	+0.65	+1.55	ND	—
BCL2L1		134	980	+1.77	—	30	2,419	+1.24	+1.21	—	—
MCL1		277	448	+2.30	—	10	5,328	+0.52	+0.93	ND	—
BCL2L2		108	1,239	+1.62	—	18	3,518	+0.92	-1.85	—	—
BAX		24	5,196	+0.55	—	6.50	6,948	+0.23	-1.54	582	—
BAK1		28	4,587	+0.66	—	3.57	9,389	-0.12	-0.05	—	—
BBC3		4.17	13,258	-0.61	—	0.25	16,979	-1.05	-0.05	—	—
PMAIP1		—	—	—	—	2.55	10,532	-0.30	+0.49	—	—
BID		21	5,771	+0.45	—	19	3,489	+0.92	-0.69	—	—
BAD		27	4,785	+0.62	—	95	797	+2.07	+1.50	—	—
BCL2L11		32	4,100	+0.75	—	0.13	18,235	-1.13	-0.97	—	—
BIRC5		7.90	10,310	-0.21	—	3.19	9,790	-0.19	+1.62	—	—
XIAP		88	1,538	+1.47	—	36	2,053	+1.37	+3.40	—	—
CFLAR		35	3,771	+0.82	—	8.54	5,965	+0.41	-2.03	—	—
CASP3		—	—	—	—	8.67	5,911	+0.42	+1.69	—	—
CASP8		—	—	—	—	0.25	16,745	-1.05	-1.14	—	—
CASP9		—	—	—	—	2.04	11,474	-0.42	-0.99	—	—
DIABLO		34	3,858	+0.80	—	25	2,789	+1.12	-0.06	—	—

Cell-cycle / CDK29 genes · max Caris Z +3.76 · max LR Z +3.08

		Caris short-read (2024 tissue)				UCLA long-read (2024 tissue)			vs GTEx	mProbe '24	Ignite RPPA
Gene	type	TPM	Rk	Z	%ile	CPM	Rk	Z	logFC	amol/μg	%
CDK1		6.53	11,198	-0.33	—	5.74	7,437	+0.16	+1.57	—	—
CDK2		6.96	10,887	-0.29	—	3.19	9,741	-0.19	+0.62	—	—
CDK4		893	123	+3.16	—	35	2,072	+1.36	+0.83	ND	—
CDK6		177	737	+1.98	—	391	195	+3.08	+6.15	272	—
CDK7		16	7,055	+0.25	—	10	5,393	+0.51	-0.54	—	—
CDK9		43	3,167	+0.95	—	3.06	9,843	-0.21	-3.36	—	—
CCNA2		0.59	18,484	-1.47	—	—	—	—	—	—	—
CCNB1		5.99	11,586	-0.39	—	8.92	5,797	+0.43	+0.21	—	—
CCND1		669	177	+2.95	20	9.43	5,598	+0.47	-0.41	ND	—
CCND2		2,055	45	+3.76	—	169	458	+2.48	+2.50	—	—
CCND3		295	416	+2.35	—	31	2,319	+1.28	-1.61	—	—
CCNE1		22	5,526	+0.49	—	0.25	16,532	-1.05	-1.26	—	—
CCNE2		19	6,238	+0.38	—	0.25	16,531	-1.05	-1.79	—	—
CDKN1A		28	4,638	+0.65	—	24	2,888	+1.09	+2.96	ND	—
CDKN1B		1,690	58	+3.62	—	36	2,036	+1.38	-0.59	397	—
CDKN2A		359	333	+2.49	74	87	879	+2.00	+4.44	—	—
CDKN2C		51	2,648	+1.09	—	71	1,078	+1.86	+3.17	—	—
RB1		339	354	+2.45	23	19	3,483	+0.93	-1.38	ND	14%
RBL1		39	3,424	+0.90	—	2.17	11,202	-0.39	+0.55	—	—
RBL2		36	3,706	+0.83	—	5.35	7,704	+0.11	-1.32	—	—
E2F1		3.50	13,942	-0.71	—	1.78	11,930	-0.48	-1.03	—	—
E2F3		2.22	15,665	-0.95	—	0.64	15,568	-0.86	-1.60	—	—
CDC25A		1.90	16,166	-1.03	—	—	—	—	—	—	—
WEE1		18	6,458	+0.35	—	1.40	12,951	-0.59	+0.58	—	—
PKMYT1		9.24	9,562	-0.11	—	0.25	16,991	-1.05	-0.09	—	—
PLK1		13	7,895	+0.12	—	1.15	13,776	-0.67	+0.32	—	—
AURKA		10	9,035	-0.04	—	0.64	15,458	-0.86	+0.39	—	—
AURKB		62	2,167	+1.22	—	1.27	13,088	-0.62	+1.19	—	—
SKP2		20	6,011	+0.42	—	14	4,258	+0.74	+3.31	—	—

Bottom line

Act-on transcriptome signals are those concordant across both platforms and supported functionally: CDK6 over-expression (both platforms + ex-vivo) and the concordant NF-κB/IκBα (NFKBIA) context, with TUBB3 a patient-specific taxane-resistance caveat. Single-platform Caris-cohort-high genes (PDGFRA, FGFR, AKT3, EGFR) require orthogonal or lineage-specific confirmation.

Proteome, summary

Two protein layers: mProbe MRM-MS targeted proteomics (2024 specimen, 282-protein panel; 2016 GPS comparator) and Ignite RPPA phospho-activation (expanded May 2026). MRM-MS non-detection is sensitivity/epitope-limited and is not proof a protein is absent.

Main proteomic findings (summary)

DNA-repair effectors

Broad HR/NHEJ/MMR/NER effector deficit by MRM-MS (ND), with PARP1 robust (1,863 amol/ug). ND = non-detection, not proven null, confirm by IHC / RAD51-foci.

Checkpoint engaged

Ignite RPPA CHK2 99 / CHK1 95 percentile, the damage checkpoint is maximally engaged; supports a single WEE1/CHK1 node.

RB / MET

RB pS780 low (14), read with the caveat that there is no formal UCLA RB IHC (informal unsigned read only); RB-loss is neither established nor excluded. MET 20 with Schürch IHC ++/+++ keeps MET a hypothesis pending CSF RNA-seq.

Chemo-relevant proteome

The 2016 to 2024 loss of ERCC1 / hENT1 / RRM1 / TOPO1 (and fixed-high TUBB3) is the chemo-sensitivity story, presented in the Chemotherapy tab.

mProbe MRM-MS, DNA-repair effector pattern

Pattern

HR / NHEJ / MMR / NER effectors (BRCA1, RAD51, RAD51C, RAD50, PRKDC, XRCC4, LIG4, MSH2, MSH6, ERCC1) below MRM detection, a broad effector deficit
PARP1 robust (1,863 amol/μg), target intact
ERCC1 127 amol (2016) → ND (2024): NER loss likely under treatment selection (panel-coverage cross-check advised)

Caveat & direction

ND = non-detection by targeted MRM-MS, not confirmed null; mRNA is normal-range for several. A genuine multi-pathway impairment is the most probable reading because it converges with an independent functional anchor (Filbin ex-vivo niraparib relative viability 0.06) and IHC-confirmed ATRX-loss/ALT, but each individual ND is soft; confirm with IHC and a RAD51-foci HR functional assay before treating as established.

Ignite RPPA, activation states

Analyte	%ile	Interpretation
CHK2 pS33/S35	99	Highest in panel, damage checkpoint maximally engaged
CHK1 pS345	95	Second highest, supports a WEE1/CHK1 checkpoint-exploitation node
ATM pS1981 / ATR pS428	30 / 0	Apical autophospho sites under-detected (ATR is present, Caris 173 TPM); downstream CHK1/2 are the reliable readouts
RB pS780 / FOXM1 pT600	14 / 4	Low RB-phospho, read with the caveat that there is no formal UCLA RB IHC (informal unsigned read only); RB-loss is neither established nor excluded
MET	20	With Schürch IHC ++/+++ keeps MET a hypothesis pending CSF RNA-seq
PD-L1 / ERK	58 / 89	Immune-evasion and MAPK-output context

Interpreting "ND" on MRM-MS protein (2024 mProbe · 2016 GPS)

Both protein platforms are targeted MRM-MS against a surrogate-peptide panel. "ND" means the surrogate peptide was below the limit of quantification on that run, not that the protein is absent. Six recognised mechanisms produce ND for an expressed protein: (1) panel-coverage gap (2016 GPS and 2024 mProbe are non-identical panels); (2) peptide selection (PTM-spanning, alternative cleavage, or non-dominant isoform); (3) FFPE crosslinking lowering recovery; (4) sub-LLOQ expression (~50–100 amol/μg threshold); (5) single-block region-of-interest sampling vs documented heterogeneity; (6) brain-tissue prep interference for membrane/low-abundance classes. ND is therefore evidence of low expression, not proof of absence.

Where this matters for Calvin

Convergent ND (mRNA-low + protein-ND + IHC-negative) is robust. Where mRNA-high disagrees with protein-ND, functional/IHC data override: FGFR1/2/3 (ND but Caris %ile 18/96/90 + Filbin ex-vivo kill), MET (ND but Schürch IHC ++/+++ and chr7q31 amplification). The BRCA1/RAD51/PRKDC/MSH2/MSH6 DDR-effector ND is treated as supportive, it converges with mRNA, ATRX-loss IHC and Filbin niraparib kill, not relied on alone. 2016→2024 "loss" calls (ERCC1 127→ND etc.) are likely given treatment selection but require panel-coverage cross-check.

Resolving a discordance

Three orthogonal routes: orthogonal proteomics (Schürch IMC / multiplex IHC, or Ignite RPPA); PRM with alternative surrogate peptides; functional ex-vivo kill (Filbin GBX63 PDX), which outweighs absolute protein number for target nomination. Weighting for target calls: mRNA + functional kill + IHC + activation state over absolute MRM-MS value.

Ignite RPPA · phospho-protein activation state (expanded panel · 15 May 2026)

A fourth orthogonal platform on the same 15 Oct 2024 UCLA Liau specimen: Reverse Phase Protein Array (RPPA) by Ignite Proteomics. RPPA quantifies activation states, phospho-site occupancy at functional residues, as percentile of a reference tumour population. Complements the abundance-based mProbe MRM-MS readout: an analyte can be ND at MRM-MS but show measurable activation on RPPA (different antibody / epitope), or abundant at MRM-MS with low activation. The expanded 35-analyte panel (15 May report) adds the DNA-damage checkpoint axis, which produces the two highest values in the entire dataset.

CHK2 99% · CHK1 95%, checkpoint hyperactivation

Expanded-panel finding. CHK2 pS33/S35 (99th) and CHK1 pS345 (95th) are the two highest activations measured. With the Axis 02 DDR-effector deficiency, this describes a checkpoint-addicted, effector-deficient tumour, direct functional rationale for CHK1/2 (prexasertib), ATR (ceralasertib, elimusertib) or WEE1 (adavosertib) inhibition on the niraparib PARP-trap trunk. ATM pS1981 30%, ATR pS428 0% (apical autophospho under-detected vs maximal downstream effectors, antibody caveat).

ERK1/2 pT202/Y204 · 89th percentile

Highest of the base panel. MEK1/2 phospho only 12%, canonical MEK→ERK does not explain it. Expanded panel adds BRAF pS445 52%: upstream MAPK more active than the MEK site implied. Strengthens direct ERK inhibitor (ulixertinib, ravoxertinib) or pan-RAF over upstream MEK blockade.

PD-L1 · 58th percentile

Directly contradicts mProbe CD274 ND and Caris mRNA Z +0.37. Functional PD-L1 surface signal is moderate-high. Re-opens checkpoint-inhibitor consideration. Cold TME (CXCL9/10 absent, Axis 07) still demands combination architecture: CXCL9-induction (IFN-γ pulse, EZH2 inhibition, oncolytic) + anti-PD-1.

RB pS780 · 14th percentile

Surprisingly low for a CDK6-amplified tumour. CDK4/6 inhibitors work by reducing pRB; if pRB is already low because RB protein is depleted (consistent with mProbe RB ND + UCLA IHC patchy positivity), drug substrate is limited. Sharpens the Axis 01 durability concern. FOXM1 pT600 at 4% reinforces low cell-cycle effector phospho activity.

MET pY1234/Y1235 · 20th percentile

Triangulates with Stanford / Schürch MET ++/+++ IHC against mProbe MET-ND. Three platforms now indicate MET is expressed and functionally active. mProbe ND is best read as a peptide-selection or FFPE-recovery artefact. Telisotuzumab vedotin (MET-DXd ADC) remains on the actionable list.

PI3K / AKT / mTOR cascade · independently confirmed low-activation

Despite AKT3 mRNA Z +4.71 and PIK3R1 Z +4.67 (Axis 05 components), RPPA reads AKT pS473 6%, AKT pT308 14%, mTOR pS2448 22%, p70S6K 12%, S6RP 5%, 4EBP1 7%. Pathway components present but cascade not constitutively active. Confirms the GSEA-based interpretation. Cycling-pressure rationale only, not a primary trunk.

HER2 / HER3 phospho both low; targets closed

HER2 total 2%, HER2 pY1248 2%, HER3 pY1289 12%. Despite ERBB3 mRNA Z +3.47, neither HER2 nor HER3 is functionally engaged at the activation level. TROP2 at 0% and AR at 3% similarly closed.

Methodology caveats for RPPA

RPPA uses antibodies against phospho-sites and total proteins, different epitope coverage from MRM-MS. (i) Antibody specificity / cross-reactivity is the dominant assay-quality variable. (ii) "% of max" is relative to a reference tumour population, not absolute quantification. (iii) Sub-cellular localisation is averaged, nuclear vs cytoplasmic vs membrane signalling cannot be separated. (iv) Single-block sample with the same heterogeneity caveats as mProbe.

Immunohistochemistry (IHC) protein panel

Targeted protein stains across UCLA surgical pathology, Schürch / PHM (Ventana), Suzhou / HUCA and Shanghai, a protein layer distinct from MRM-MS.

Source	Positive / informative	Negative / not detected
UCLA surgical path · SSW-24-26425 · 15 Oct 2024	GFAP+ · p53+ · ATRX loss · H3 G34R/V+ · Ki-67 20–40% · synaptophysin patchy+ · INSM1 rare+ · PD-L1 weak-focal	IDH1 R132H − · HER2 − · PD-1 − ; cold immune TME (CD3/CD8 sparse, CD4 reduced, CD45/CD68 low)
Schürch / PHM · Ventana · Mar 2026	MET ++/+++ · EGFR +++ · B7-H3/CD276 ++ · CD147/BSG +++ · PDGFRβ ++/+++ (vasc) · FGFR3 (+) · HER3 weak · Nectin-4 weak · NY-ESO-1 v.weak	n.d.: CTLA-4 · LAG3 · TROP2 · MUC1 · ROS1 · CEACAM5 · Claudin-18 · PRAME · CSPG4 · mesothelin · PDGFRα · GPC3 (re-review flag)
Suzhou / HUCA · Feb 2026	EGFR H-score 190 (moderate)	B7-H3 H-score 3 (weak) · IL13Rα2 0 · HER2 0
Shanghai (tracking sheet only)	PSMA weak+ (~40%) · AXL weak+ (~20%)	CR-1 − · IL-1RAP −

IHC reconciliation, decision-critical

No formal UCLA RB IHC exists, only an informal, unsigned "patchy" read with no primary report. RB protein status is effectively unproven; the CDK4/6i durability question is not reassured by IHC. Gate on a formal RB IHC plus CSF RNA-seq.
No MMR IHC was ever performed (MLH1 / PMS2 / MSH2 / MSH6 never stained). The proteomic MMR-ND is therefore unresolved at protein level; only molecular MSI-stable / TMB-low support intact MMR. The temozolomide-hypermutation question cannot be closed on IHC, an MMR IHC + MSI/PCR + SBS11 signature is still required.
Discordances: B7-H3/CD276 Schürch ++ vs Suzhou weak (H-score 3) on the same block, material for any B7-H3 CAR-T rationale. EGFR is strongly IHC-positive (Schürch +++, Suzhou H-190) but not amplified on NGS and stains adjacent normal brain, most consistent with assay/background, not an EGFR-amplified target.
Primary reports absent (referenced second-hand only): Stanford GD2, the original 2016 MGH neuropathology, and the Shanghai / Mayo panels.

Bottom line

The robust, act-on proteomic readouts are the checkpoint activation (CHK2 99 / CHK1 95, direct RPPA measurement, supporting a single WEE1/CHK1 node) and PARP1-intact in an ATRX-loss/ALT background. The DDR-effector ND pattern is supportive of, not proof of, multi-pathway repair deficiency, confirm orthogonally (IHC, RAD51-foci).

Chemotherapy & chemo-sensitivity

Calvin's 2024 tumour is the product of ~10 years of treatment selection. This covers the 2016 to 2024 chemo-sensitivity protein evolution, which conventional classes are now exhausted, and the one conventional agent that still has a clear untested rationale and by which route.

2016 to 2024 chemo-sensitivity protein evolution

GPS Cancer MRM-MS (Feb 2016) and mProbe MRM-MS (Oct 2024) quantify targeted chemo-sensitivity proteins. Caris short-read (N=19,168) and UCLA long-read (N=19,952) provide Oct 2024 mRNA. Where protein is absent but transcript persists the discordance is noted separately. Protein values: amol/µg. Z-scores: population-normalised (Caris log-mean 1.07 SD 0.59; long-read 0.73 SD 0.60).

The 10-year remodelling: what is present, what is lost, and what that means

Four proteins quantified in 2016 are below detection in 2024: ERCC1 (127→ND), hENT1/SLC29A1 (212→ND), RRM1 (610→ND), TOP1 (1,090→ND). Three of these genes retain mRNA in 2024 — ERCC1 Caris Z +1.66, SLC29A1 Z −0.05, RRM1 Z +0.18. Transcript persists but protein is gone. The most parsimonious explanation is post-transcriptional silencing or protein instability driven by treatment selection over 10 years, not gene-level deletion (no copy-number loss for any of these loci at Caris WES). Persistent mRNA does not restore enzymatic function. TUBB3 is the inverse: protein was high in 2016 (15,400 amol) and unchanged in 2024 (15,713 amol), with concordant RNA elevation on both platforms (Caris Z +1.65, LR Z +2.19, logFC +1.45 vs normal brain). This is a constitutive expression pattern predating any therapy. The NER-deficient state emerging from ERCC1 protein loss creates a platinum sensitivity that has not been tested in 10 years of treatment. Nucleoside-transport and reductase loss makes gemcitabine and nucleoside analogues mechanistically futile. Taxane resistance was present before treatment began.

Gene	Drug-class function	2016 amol/µg	2024 amol/µg	Caris Z mRNA	LR Z mRNA	logFC vs brain	Protein–RNA concordance and drug-class consequence
ERCC1	NER endonuclease; canonical platinum-resistance effector	127	ND	+1.66	+1.19	−1.22	mRNA present, protein absent (discordant). NER-deficient at the functional (protein) level → carboplatin / cisplatin sensitivity predicted. Carboplatin untried in 10 years of treatment.
hENT1 (SLC29A1)	Equilibrative nucleoside transporter; gemcitabine / cytarabine cellular uptake	212	ND	−0.05	−0.30	−0.68	mRNA near population median, protein ND (discordant). Transporter absent → gemcitabine and IT-cytarabine cannot enter cells regardless of extracellular concentration.
RRM1	Ribonucleotide reductase M1; gemcitabine molecular target; high expression = resistance	610	ND	+0.18	+0.32	+1.31	mRNA mid-range, protein ND (discordant). Target lost — reinforces gemcitabine futility independently of hENT1-loss.
TOP1 (TOPO1)	Topoisomerase I; molecular target of irinotecan and topotecan	1,090	ND	—	—	—	Protein lost; TOP1 not detected in mRNA panel. Target absent — topo-I-poison activity reduced. IT-topotecan is used empirically for LMD; protein-ND provides context but does not eliminate empirical use.
TUBB3	Class III β-tubulin; taxane and vinca-alkaloid resistance	15,400	15,713	+1.65	+2.19	+1.45	Protein stable across both timepoints; RNA concordantly elevated on both platforms and above normal brain (logFC +1.45). Constitutive expression at every molecular level — taxane / vinca resistance was present before first-line therapy and has not changed.
EGFR	RTK (chemo context only; no direct chemo-sensitivity role)	~299	~299	+3.24	−0.05	+0.07	Short-read mRNA high (Z +3.24) but long-read near-normal (Z −0.05) and Ignite RPPA 7% activation — platform discordance, no chemo implication. Protein stable. IHC strongly positive (UCLA surgical path), consistent with tumour lineage not amplification.
MGMT	O6-methyltransferase; repairs TMZ-induced O6-methylguanine lesions; methylated promoter silences expression	—	low	+0.59	+1.20	−0.72	Protein low; 15% promoter methylation. mRNA mid-range. In an MMR-intact tumour, MGMT methylation would predict TMZ benefit. MMR-effector proteins lost (MSH2/MSH6 ND at protein, SBS11 signature at 2024) — TMZ in this setting drives hypermutation not cytotoxicity.

Therapeutic considerations

Exhausted / futile classes

Temozolomide, exclude: MMR-protein loss converts TMZ from cytotoxic to a hypermutagen (SBS11); MGMT-methylation alone is insufficient (see Drug screen / Candidate menu)
CCNU (lomustine), 5 cycles, progressed
Taxanes / vinca, TUBB3 fixed-high since 2016: intrinsic resistance, not worth re-using
Gemcitabine, hENT1 (uptake) and RRM1 (target) both lost: mechanistically futile
IT-cytarabine, hENT1-loss similarly undermines nucleoside-analogue uptake

Carboplatin, clearest untested conventional rationale

Carboplatin was not used over the preceding 10 years. ERCC1, the NER protein that conventionally drives platinum resistance, is lost (127 to ND), which is expected to sensitise to carboplatin. Caveats: Grade-4 lymphopenia (ALC 0.21) makes systemic carboplatin marrow-risky; intrathecal carboplatin via Ommaya is reported for glioma leptomeningeal disease and limits systemic exposure. This is the conventional agent with the clearest mechanism-based untested basis here, but it requires functional confirmation (ideally on Calvin's own cell line) before adoption. Cross-reference the Drug-screen carboplatin band and the Candidate-menu row.

Bottom line

Re-using exhausted classes (TMZ, CCNU, taxanes, gemcitabine) is not worthwhile on the proteomic and clinical evidence. The single conventional chemotherapy with a clear, mechanism-based, untested rationale is carboplatin (ERCC1-loss), preferably intrathecal via Ommaya given marrow constraints, gated on functional confirmation. An adjunct option, not a backbone, it does not displace the CDK4/6 + replication-stress spine.

Genomics, integrated landscape (all sources)

An H3.3 G34R diffuse hemispheric glioma: truncal backbone of H3-3A G34R + TP53 R273C + ATRX-loss/ALT and a truncal CDK6 amplification, now leptomeningeal, with a subclonal recurrence-emergent PDGFRA Y288C; several copy-number and protein-loss calls are low-reliability and flagged as such. Integrated genomic landscape across all layers: tumour-tissue WES (Caris MI Tumor Seek, 15 Oct 2024), the UCLA Pan-Cancer Solid Tumor NGS panel (1,080-gene, matched tumour/normal, reported 23 Dec 2024), CSF cell-free DNA (Belay, Shanghai Decode, SiMSen-Seq, OncoIncytes; 2026), long-read RNA fusions (JAFFAL), copy number, germline / mismatch-repair / tumour mutational burden, and the established genomics of H3 G34-mutant DHG.

Integrated genomic landscape, multi-source

Truncal, tissue WES + CSF concordant

H3-3A (H3F3A) G34R, defining driver
TP53 p.R273C, truncal, high CSF VAF/MM/mL; contact mutant
ATRX loss → ALT phenotype
TMB 1.0 → 4.31 (rising); MGMT methylated (PyroSeq); MMR-protein loss (MSH2/MSH6)

Subclonal / LMD-emergent, CSF ctDNA

PDGFRA p.Y288C, absent in 2024 tissue; emerged in CSF 2026; resistant to the PDGFR TKIs tested by Ip 2018 (avapritinib untested, plausibly but not proven resistant); subclonal/regressing
NF1 R156C, transient 28-Feb subclone (cleared)
SiMSen 16-variant personalised panel (MTIF3, CCR3, MICAL2, ALK, WT1…)

Structural, long-read fusions

CDK6:VIPR2 (292 reads) + CDK6:KRIT1 / FAM133B:CDK6, chr7q amplicon-associated rearrangements that corroborate CDK6 amplification, not oncogenic fusions
CYP46A1:BRAF, 2 reads, single-platform; hypothesis-only, unconfirmed
XPO1:PPP4R3B, single call, likely passenger; does not establish XPO1 dependency
No NTRK / FGFR / FGFR3-TACC3 fusion (screened, see Candidate menu)

Copy number, provenance-flagged

CDK6 amplification, 2024 tissue Caris CNA; not re-confirmed in recent CSF LP-WGS
Belay Ascent LP-WGS focal CNV (RB1-del / MET-amp / MDM4-amp), reliability LOW, see ctDNA tab
CDKN2A/9p loss reported (same low-reliability caveat)

Germline / DDR

ATRX-loss → ALT (replication-stress / PARP-relevant)
Germline DDR panel, pending (gating test)
RAD51-foci HR functional assay, recommended (mProbe DDR-ND unreliable)

Tumour-type biology, H3 G34R DHG

Arises from a ventral-forebrain interneuron / OPC-competent progenitor; near-universal TP53 + ATRX co-mutation (matches this case); PDGFRA amp/activation a recurrent co-driver; CDKN2A loss common; MGMT often methylated. A distinct entity from H3 K27M, the ONC201 biomarker does not transfer. Hallmark OPC↔astrocyte↔mesenchymal lineage plasticity underlies therapy escape (see 6-week blocks item 4–5).

Variant × platform heatmap · cross-platform reconciliation

Each cell = VAF % at that platform. Colour intensity scales with VAF (red high → pale → blue low). Empty cells = not reported. AMP = copy-number amplification. VUS = variant of uncertain significance. Same 2024-10-15 tissue specimen for Caris MI and UCLA Pan-Cancer NGS, different panels and report dates, but one biological sample. The 2016 GPS Cancer column (NantOmics S16-12505 R frontal · reported 8 Jun 2016) confirms H3.3 G34R and TP53 R273C were already present 9 years earlier (33/152 and 32/40 reads respectively), single monoclonal disease across the entire course. Click any gene row for the entity card.

		TISSUE · CHRONOLOGICAL			CSF · LONGITUDINAL
Gene · variant	CLASS	GPSJUN 2016	CARISOCT 2024	UCLAOCT 2024	PREDICINE24 JUL 2025	SHANGHAI5 AUG 2025	BELAY5 FEB 2026	SHANGHAI5 FEB 2026	ONCOINCYTES6 FEB 2026	SIMSEN26 FEB 2026	SHANGHAI28 FEB 2026	SHANGHAI18 MAR 2026	SHANGHAI25 MAR 2026	SHANGHAI27 MAR 2026	BELAY13 APR 2026
TP53 p.R273C	TRUNCAL	80%	45%	73%	1.89%	1.89%	79.2%	ND	93%	51.16%	20.13%	12.9%	9.62%	5.31%	24.6%
MYH11 p.M741Rfs*10	TRUNCAL	ND	23%	39%	ND	ND	ND	ND	ND	ND	ND	ND	ND	ND	ND
ALK c.2487+1G>A	TRUNCAL	ND	7%	31%	ND	ND	ND	ND	ND	ND	ND	ND	ND	ND	ND
H3-3A p.G35R (G34R)	TRUNCAL	21.7%	17%	27%	1.03%	1.03%	25.4%	ND	ND	ND	1.9%	5.78%	4.5%	2.33%	6.4%
CREB3L2 p.K176E	TRUNCAL-CSF	ND	55%	ND	ND	ND	ND	ND	ND	ND	ND	ND	ND	ND	ND
HGF p.R393H	TRUNCAL-CSF	ND	51%	ND	ND	ND	ND	ND	ND	ND	ND	ND	ND	ND	ND
KAT6A p.D1317V	TRUNCAL-CSF	ND	51%	ND	ND	ND	ND	ND	ND	ND	ND	ND	ND	ND	ND
EME1 p.Q394*	TRUNCAL-CSF	ND	51%	ND	ND	ND	ND	ND	ND	ND	ND	ND	ND	ND	ND
TAT p.R297*	TRUNCAL-CSF	50.6%	ND	ND	ND	ND	ND	ND	ND	ND	ND	ND	ND	ND	ND
PTCH1 p.K353R	TRUNCAL-CSF	ND	50%	ND	ND	ND	ND	ND	ND	ND	ND	ND	ND	ND	ND
RCAN1 p.A34E	TRUNCAL-CSF	ND	50%	ND	ND	ND	ND	ND	ND	ND	ND	ND	ND	ND	ND
AXIN2 p.Y550D	TRUNCAL-CSF	ND	49%	ND	ND	ND	ND	ND	ND	ND	ND	ND	ND	ND	ND
TOP3A p.M729V	TRUNCAL-CSF	ND	49%	ND	ND	ND	ND	ND	ND	ND	ND	ND	ND	ND	ND
CDK12 p.P530A	TRUNCAL-CSF	ND	48%	ND	ND	ND	ND	ND	ND	ND	ND	ND	ND	ND	ND
MAML2 p.S199L	TRUNCAL-CSF	ND	48%	ND	ND	ND	ND	ND	ND	ND	ND	ND	ND	ND	ND
SGK1 p.E392Q	TRUNCAL-CSF	ND	48%	ND	ND	ND	ND	ND	ND	ND	ND	ND	ND	ND	ND
SLC6A4 p.E2G	TRUNCAL-CSF	44.2%	ND	ND	ND	ND	ND	ND	ND	ND	ND	ND	ND	ND	ND
ATRX p.E2270del	TRUNCAL-CSF	ND	44%	ND	ND	ND	det	ND	ND	ND	18.64%	11.84%	12.4%	6.46%	23.5%
AXIN1 c.2587_*1dupTGAT	TRUNCAL-CSF	ND	43%	ND	ND	ND	ND	ND	ND	ND	ND	ND	ND	ND	ND
MYO6 p.R295C	TRUNCAL-CSF	41.9%	ND	ND	ND	ND	ND	ND	ND	ND	ND	ND	ND	ND	ND
COL1A1 p.G1022V	TRUNCAL-CSF	ND	ND	38%	ND	ND	ND	ND	ND	ND	ND	ND	ND	ND	ND
LPP p.P400L	TRUNCAL-CSF	28.9%	ND	ND	ND	ND	ND	ND	ND	ND	ND	ND	ND	ND	ND
PDGFRA p.Y288C	TRUNCAL-CSF	ND	ND	ND	ND	ND	28.5%	ND	CTC 38.7	ND	2.93%	1.95%	0.74%	1.04%	0.7%
BBC3 p.G193R / p.E158	TRUNCAL-CSF	ND	ND	ND	ND	ND	ND	ND	ND	25.33%	ND	ND	ND	ND	ND
RAD50 p.E723Gfs*5	SUBCLONAL / VUS	ND	ND	ND	2.54%	ND	ND	ND	ND	ND	ND	ND	ND	ND	ND
GATA3 p.C375R	SUBCLONAL / VUS	ND	INT	ND	0.82%	ND	ND	ND	ND	ND	ND	ND	ND	ND	ND
FAT1 p.D864G	SUBCLONAL / VUS	ND	ND	ND	0.81%	ND	ND	ND	ND	ND	ND	ND	ND	ND	ND
BAP1 p.H141R	SUBCLONAL / VUS	ND	ND	ND	0.69%	ND	ND	ND	ND	ND	ND	ND	ND	ND	ND
ATM p.H1951R	SUBCLONAL / VUS	ND	ND	ND	0.65%	ND	ND	ND	ND	ND	ND	ND	ND	ND	ND
BRCA1 p.Q1556R	SUBCLONAL / VUS	ND	ND	ND	0.64%	ND	ND	ND	ND	ND	ND	ND	ND	ND	ND
MPL p.W515L	SUBCLONAL / VUS	ND	ND	ND	0.23%	ND	ND	ND	ND	ND	ND	ND	ND	ND	ND
POLD2 p.V292M	MIXED	ND	23%	ND	ND	ND	ND	ND	ND	ND	ND	ND	ND	ND	ND
ERG p.R130H	MIXED	ND	6%	ND	ND	ND	ND	ND	ND	ND	ND	ND	ND	ND	ND

VAF≥7045–7025–4510–252–10<2ND

What the genomic layer changes

Key genomic findings, provenance & cross-platform status

H3-3A G34R, TP53 R273C, ATRX-loss, truncal; concordant across tissue WES and CSF (ATRX additionally IHC-confirmed).
PDGFRA p.Y288C, tissue-WES-negative, CSF-positive (28.5%); subclonal / regressing. A Y288C-directed PDGFR-TKI is not well-supported (detail in Pre-therapy review item 3); downstream PI3K/mTOR or MEK is better-grounded.
CDK6 amplification, 2024 tissue Caris CNA; corroborated by the long-read CDK6:VIPR2 amplicon rearrangement and cross-platform-concordant over-expression; not re-confirmed in recent CSF LP-WGS.
RB1, only a Belay LP-WGS focal-deletion call (low-reliability); Ignite RB pS780 14% and mProbe RB ND, and there is no formal UCLA RB IHC (informal unsigned read only). RB protein status is effectively unproven, confirm with a formal RB IHC plus CSF RNA-seq before CDK4/6i decisions.
MET, chr7q31 amplification only on the same low-reliability Belay run; Schürch IHC ++/+++ and Ignite 20%. A hypothesis pending CSF RNA-seq, not DNA-confirmed.
CDK12 p.P530A, constitutional, concordant across Belay and the Decode runs; relevant to HR-transcription / DDR-effector status.

Flags for the MDT

MGMT discordance: Caris PyroSeq = Methylated (TMZ-benefit Level 2) vs NeoGenomics 15% borderline. Resolve which assay/threshold governs TMZ-backbone decisions.
EME1 Q394* nonsense, adds a structure-specific-endonuclease LOF to the DDR catastrophe; reinforces the checkpoint-inhibition rationale (Ax02).
HGF R393H, MET-ligand variant; possible context for the Ignite MET activation.
BLMH:NF1 fusion, candidate contributor to the Ignite ERK-high / MEK-low MAPK pattern (Ax05).
KAT6A D1317V (~51%), possible germline chromatin-modifier variant, not previously actioned.

Orthogonal DNA validation, UCLA Pan-Cancer Solid Tumor NGS panel

UCLA Molecular Diagnostics "Pan-Cancer All v1", 1,080-gene hybrid-capture NGS with a matched normal (somatic calls only), specimen 24RR-MD05142, same surgical resection (SSW-24-26425, block D1) as the Caris WES (block C1), reported 23 Dec 2024. An independent assay on the same tumour.

Finding	UCLA Pan-Cancer (VAF)	Caris WES	Concordance
H3-3A p.G34R (reported p.G35R, new HGVS)	27%	17%	CONCORDANT, truncal driver
TP53 p.R273C	73%	45%	CONCORDANT, truncal
ATRX loss	IHC loss (panel: no SNV called)	E2270del + IHC loss	CONCORDANT (IHC)
TMB / MSI	4.31 mut/Mb · MSS	~5 mut/Mb · MSS	CONCORDANT
1p/19q · chr7 gain / chr10 loss	not detected	not detected	CONCORDANT
MGMT promoter methylation	DETECTED (NeoGenomics addendum, qualitative)	methylated	CONCORDANT (direction)
ALK c.2487+1G>A (splice-donor)	31%	not reported	UCLA-unique, unclear significance
MYH11 p.M741Rfs*10 · COL1A1 p.G1022V	39% · 38%	not reported	UCLA-unique, likely passengers
CDK6 amplification	not assessable (CNVs not reported by design)	amplified (WES CNA)	not a discordance, method limitation

Interpretation

The UCLA panel independently confirms every truncal call (H3-3A G34R, TP53 R273C, ATRX-loss, TMB-low/MSS, MGMT-methylated) on a matched tumour/normal design, strong orthogonal validation of the genomic backbone; VAF differences (TP53 73 vs 45%, H3 27 vs 17%) reflect adjacent-block tumour-content, not discordance. Caveats: (1) the panel does not report CNVs by design, so it neither confirms nor refutes the WES CDK6 amplification, not a true discordance and consistent with the CDK6 provenance note above. (2) ALK c.2487+1G>A is a splice-donor variant of unclear oncogenic significance in glioma with no determinable fusion partner on a DNA panel; together with the long-read finding of no canonical ALK fusion it does not establish ALK-targetable activation (does not qualify for ALK-directed therapy on current evidence, relevant to the Ignite ALK/ROS1 46% signal). (3) MYH11 / COL1A1 are not established glioma drivers, most likely passengers. No directly contradictory call exists between the two assays.

Somatic SNV / indel · tumour DNA

Gene	Variant	VAF	Class	Axis	Note
H3F3A	p.G35R (G34R) · Ex2 c.103G>A	17%	Pathogenic	Ax06	Founding histone driver. Low VAF reflects diluting normal brain in an infiltrative specimen, not subclonality.
TP53	p.R273C · Ex8	45%	Pathogenic	Ax03	Trunk variant. Matches the SiMSen CSF trunk (51–56% VAF). APR-246 / HSP90-client logic in Ax03.
ATRX	p.E2270del · Ex31	44%	VUS	Ax02·Ax06	ATRX alteration corroborating the IHC ATRX-loss / ALT phenotype. Also seen as ATRX:MAGT1 fusion.
CDK12	p.P530A · Ex2	48%	VUS	Ax02	~50% VAF + no informative somatic ctDNA (NorthStar) = probable germline. CDK12 regulates BRCA1/RAD51 transcription, mechanistic root of the DDR-effector collapse.
POLD2	p.V292M · Ex8	23%	VUS	Ax02	Polymerase-δ subunit, candidate contributor to the mutator phenotype (with MMR loss).
KAT6A	p.D1317V · Ex17	51%	Unclassified	Ax06	~50% VAF, possible germline. H3K23 acetyltransferase; chromatin-state relevance not previously actioned.
EME1	p.Q394* · Ex6	51%	Unclassified	Ax02	Nonsense in a structure-specific endonuclease (MUS81-EME1). LOF here further cripples replication-fork / HJ resolution, adds to the DDR-catastrophe and checkpoint-addiction picture.
HGF	p.R393H · Ex10	51%	Unclassified	Ax04	MET ligand variant. Relevant to the Ignite MET pY1234/1235 20% + Schürch IHC ++/+++, possible ligand-context contribution to MET activation.
ALK	c.2487+1G>A splice · Ex14	7%	Unclassified	Ax04	Subclonal (7% VAF) splice-site variant. Not a canonical activating ALK lesion; no ALK fusion on long-read.
PTCH1	p.K353R · Ex7	50%	VUS	—	Hedgehog pathway; ~50% VAF possible germline. Not currently an axis.
AXIN1 · AXIN2	AXIN1 dup Ex11 (43%) · AXIN2 p.Y550D (49%)	43/49%	VUS / Uncl.	—	Two Wnt-destruction-complex variants, Wnt-axis perturbation not currently modelled.
Others	CREB3L2 K176E · MAML2 S199L · MYH11 M741fs · RCAN1 A34E · SGK1 E392Q · TOP3A M729V · ERG R130H	6–55%	Unclassified	—	Lower-priority unclassified variants. TOP3A (topoisomerase III-α) and MAML2 (NOTCH co-activator) flagged for cross-reference.

Copy number · genomic signatures · HLA · key negatives

Copy number

CDK6 Amplified

The only reported CNA. Hard confirmation of the Axis 01 truncal CDK6 amplification asserted from expression + fusion data. No CDKN2A/B deletion, no EGFR / ERBB2 amplification.

Signatures

TMB 5 Mt/Mb (Low)
MSI Stable
LOH Low · 4% of segments
MGMT Methylated (PyroSeq) · TMZ-benefit Level 2

HLA class I

A*02:06 · A*11:01
B*48:03 · B*18:01
C*07:04 · C*08:01

A*02:06 confirmed, supports the PRAME / HLA-A*02-restricted TCR-T pathway.

Decision-relevant negatives

PDGFRA: tissue-negative (Oct 2024 WES), but see ctDNA tab. PDGFRA p.Y288C activating mutation IS present in CSF ctDNA (28.5%, recurrence-emergent). The WES negative reflects the Oct 2024 tissue only.
IDH1/2 WT · TERT promoter WT
BRAF mutation not detected (Caris); no BRAF fusion called by Caris
NTRK / RET / MYB / MYBL1 / EGFR fusions not detected
CDKN2A/B no deletion · NF1 / PTEN / SETD2 WT

Caris RNA fusions · onco-relevant subset (≈90 total, all "Unclassified Fusion Detected")

Caris short-read fusion calls. None are classified pathogenic, but several recur with the JAFFAL long-read set or touch axis genes. The recurrent GNAS 3′-partner cluster corroborates the Axis 09 GNAS overexpression (Z +5.30).

Fusion	Axis	Cross-reference / relevance
CDK6:VIPR2 · CDK6:LINC00689 · ANKIB1:VIPR2 · VPS50:LINC00689	Ax01	chr7q amplicon-associated rearrangement cluster, partners share chr7q (CDK6 7q21, VIPR2 7q36). These corroborate the truncal CDK6 amplification but are not classic oncogenic CDK6-activating fusions; the actionable lesion is CDK6 amplification / over-expression with intact RB, not a fusion (see long-read interpretation below).
EZR:QKI · PRKN:QKI	Ax06	Two independent QKI fusions (matches long-read PARK2:QKI / MAP3K4:QKI). Recurrent QKI (6q TSG / oligo-chromatin regulator) disruption.
AKAP13:GNAS · ATP1A2:GNAS · GPSM2:GNAS · SF1:GNAS · TOLLIP:GNAS	Ax09	Five different 5′ partners converging on a shared GNAS 3′ end, the classic signature of a recurrent-3′-partner / promiscuous-hub artifact (highly-expressed GNAS), not evidence of GNAS activation. GNAS activation is not an established glioma driver, interpret as artifact.
ATRX:MAGT1	Ax02·Ax06	ATRX fusion alongside the ATRX E2270del SNV and IHC ATRX-loss, convergent ATRX disruption.
UPF2:SMARCA4 · NSD1:ZNF346 · NSD2:KDM2A	Ax06	SWI/SNF (SMARCA4) and NSD1/NSD2 chromatin-machinery fusions, consistent with the G34R chromatin-remodelling axis.
EIF4A1:AKT2 · MYO1D:STAT3 · BLMH:NF1	Ax05	AKT2, STAT3 and NF1 fusion partners in cascade genes. BLMH:NF1 is notable given the Ignite ERK 89%, NF1 disruption de-represses RAS/MAPK.
XPO1:PPP4R3B · ZHX3:TOP1 · SF3B1:APP	—	Single calls; likely passengers. No recurrent oncogenic XPO1 fusion is established (XPO1 oncogenicity is driven by point mutation / over-expression); XPO1:PPP4R3B does not establish XPO1 dependency and does not independently support a selinexor rationale.

JAFFAL long-read RNA fusions, orthogonal validation

Long-read fusion calls (2,007 total; 144 high-confidence; 38 in-frame), shown alongside the Caris short-read set. Most are structural / amplicon / passenger events, not oncogenic drivers; see the weighting rubric below.

Fusion	Reads	Frame	Caris corrob.	Interpretation (literature-grounded)
CDK6:VIPR2	292 (+10)	partial	yes	chr7q amplicon-associated rearrangement, not an oncogenic fusion. Corroborates CDK6 amplification (the actionable lesion, RB-intact → CDK4/6i rationale stands), do not report as a "CDK6 fusion driver".
CDK6:KRIT1	20	in-frame	—	Second chr7q21 CDK6 rearrangement (CDK6 and KRIT1 both 7q21), reinforces a CDK6 structural-rearrangement / amplicon locus, not a kinase-activating fusion.
FAM133B:CDK6	2	in-frame	—	Third low-read CDK6 partner, CDK6 as a multi-partner chr7q rearrangement hub; amplicon noise.
CYP46A1:BRAF	2	in-frame	no	hypothesis only Theoretically actionable only if the transcript loses BRAF exons 1–8 (autoinhibition) and retains an intact in-frame kinase domain (KIAA1549:BRAF paradigm). At 2 reads, single-platform, not Caris-confirmed, requires RT-PCR / DNA-breakpoint confirmation before any RAF/MEK inference; not a driver on current data.
EZR:QKI · PARK2:QKI · MAP3K4:QKI	2 each	1 in-frame	yes (EZR/PRKN/MAP3K4:QKI)	Recurrent QKI involvement across both platforms → likely 6q QKI tumour-suppressor loss-of-function (oligodendroglial / RNA-binding TSG), not a kinase driver. No QKI:NTRK present (NTRK actionability does not apply). Not directly druggable.
ATRX:MAGT1	—	—	yes	ATRX and MAGT1 adjacent on chrX → local intragenic rearrangement disrupting ATRX (a second hit), not a gene fusion. Convergent with ATRX E2270del + IHC loss → supports ALT-positivity (ATR-inhibitor-relevant biology, not a targetable fusion).
ROS1 / ALK	0 (HC set)	—	—	Negative. No canonical ROS1/ALK fusion despite Ignite ROS1 pY2274 46%, the ROS1 signal is not fusion-driven; does not qualify for lorlatinib (NCT06333899) on current evidence.

MDT weighting rubric for fusion calls

Read count: ≥100 clean spanning reads = worth pursuing; 2–20 reads = hypothesis-only, insufficient alone to call a driver or change therapy.
Cross-platform: calls corroborated on both Caris (short-read) and JAFFAL (long-read) carry far more weight; single-platform + low-read + in-frame-only is the weakest tier (e.g. CYP46A1:BRAF), confirm orthogonally before clinical use.
Same-chromosome-arm partners (CDK6/KRIT1/VIPR2 on 7q; ATRX/MAGT1 on Xq) ⇒ presume local rearrangement / readthrough / amplicon artifact unless a known oncogenic mechanism is demonstrated.
Functional-domain test: a fusion is a plausible driver only if it retains an intact in-frame 3′ effector/kinase domain and loses autoinhibition (BRAF paradigm).
Promiscuous 3′ hub (one 3′ gene, many 5′ partners, GNAS here) ⇒ artifact until proven otherwise.

Net: no fusion in this case should be presented as a confirmed oncogenic driver. The truncal actionable biology is CDK6 amplification (RB-intact → CDK4/6i) and convergent ATRX inactivation / ALT; CYP46A1:BRAF is the only call with theoretical actionability and is currently unconfirmed and underpowered.

ctDNA

Longitudinal CSF circulating-tumour-DNA monitoring across the multiple assay platforms. CSF is the monitored compartment because the disease is leptomeningeal and blood sheds little tumour DNA.

Monitoring programmes

Programme	Measures	LOD / sensitivity	Role in decisions	Reliability here
SiMSen-Seq personalised (16-variant, Ommaya)	Absolute molecules/mL per patient-specific variant (UMI)	single-molecule	Dilution-robust clonal burden, the discriminating metric	Anchor
Belay Summit 2.0 (520-gene) + Ascent LP-WGS	Panel VAF + focal CNV / arm aneuploidy	VAF 0.30%	Variant detection; CNV calls are NOT decision-grade	VAF moderate · LP-WGS CNV low (see flag)
Shanghai Decode (1200-gene)	Panel VAF	2%	Cross-platform variant corroboration	Moderate
OncoIncytes CTC + ctDNA	CSF circulating tumour cells + ctDNA	—	Orthogonal confirmation (e.g. PDGFRA Y288C CTC 38.7%)	Supportive
CSF cytology	Malignant cells	—	LMD confirmation (low molecular resolution)	Adjunct
Amino-acid PET + RANO MRI (brain + whole spine)	Metabolic + anatomic burden	—	Non-molecular orthogonal response axis	Anchor (imaging)

Interpretation logic

VAF is dilution-confounded

Multiple CSF-directed interventions (intrathecal CAR-T, oncolytic virus) sat between draws and inflate the non-tumour cfDNA denominator, so VAF ratios cannot be read as response. Apparent declines within a 2–5× band are not established change.

Absolute MM/mL is the discriminator

SiMSen absolute burden: total 11,592 MM/mL on 26 Feb baseline (trunk TP53 R273C 2,095 MM/mL) → 48,989 MM/mL at CSI Day 7 (acute radiation release, +4.23×) → cfDNA 2.80 ng at Day 16 (near baseline; full per-variant MM/mL pending). Absolute counts, not VAF, adjudicate clonal change.

SiMSen-Seq personalised CSF, weekly serial kinetics

16-variant patient-specific panel sampled via Ommaya, reporting absolute molecules/mL per variant (less affected by cfDNA dilution than VAF). Timeline: 26 Feb baseline → Day 7 of CSI; Day 16 / Day 29 pending.

Trunk TP53 R273C 51.16% / 2,095 MM/mL at the 26 Feb baseline (total 11,592 MM/mL, cfDNA 1.50 ng). At CSI Day 7 the total rose to 48,989 MM/mL with cfDNA up 18×, a mixed signal: interval growth plus an acute apoptotic CSI release (the disproportionate cfDNA flood is a release signature). By Day 16 cfDNA had returned to ~baseline (2.80 ng); per-variant MM/mL for Day 16 / 29 are pending. Imaging improved over the same window (30 Apr / 1 May MRI, first non-progression since LMD onset).

Multi-platform variant & CNV detail

Per-platform variant and copy-number detail across the CSF assays.

PDGFRA, tissue-negative, CSF-positive

PDGFRA p.Y288C, activating, CSF-positive

The Caris WES (Oct 2024 tissue) reported PDGFRA mutation not detected. The CSF ctDNA monitoring detects PDGFRA p.Y288C (Class II, Activating) at 28.5% VAF (5 Feb 2026, Belay + Shanghai Decode; OncoIncytes CSF-CTC 38.7%). It is a recurrence / LMD-compartment-emergent activating mutation, exactly the Funato 2020 pattern (PDGFRA selected at recurrence). Avapritinib rationale is not well-supported for this variant (unproven, not excluded): Y288C is an ER-trapped extracellular Ig-D3 mutant whose oncogenicity is ligand- and kinase-conformation-independent, characterised as resistant to the PDGFR TKIs tested by Ip 2018 (Nat Commun, PMID 30389923), and the D842V→avapritinib type-I logic does not transfer. Avapritinib itself was not tested in Ip 2018 (the authors recommended evaluating it), so avapritinib futility here is mechanistically plausible but not directly demonstrated. Better-grounded targets are downstream PI3K/mTOR + MEK. See Pre-therapy review item 3 (variant-specific pharmacology) and item 2 (this clone is subclonal/LMD-emergent/regressing, not truncal).

Germline confirmed (concordant all platforms)

CDK12 p.P530A, germline VUS (constitutional) across Belay + 4 Shanghai Decode runs. Strengthens the Ax02 hypothesis that a germline CDK12 hit is the transcriptional root of the HR/DDR-effector collapse.
AXIN2 p.Y550D, germline VUS (constitutional).

Signatures · MGMT resolution

MSI MSS, concordant all platforms
TMB 13.08 → 0.93 Muts/Mb (declining, TMB-Low)
MGMT Methylated (Vantage qPCR+HRM, NCCN Tier 1A), but MMR-loss + TMB trajectory temper TMZ-rechallenge enthusiasm. Resolves the Caris-vs-NeoGenomics discordance: methylated, with a caveat.
HLA heterozygous all loci, A*02:06 confirmed (PRAME TCR-T)

Belay Ascent LP-WGS focal CNV (13 Apr 2026), reliability flag

Reliability assessment

The chr7p/EGFR arm gain present on the first Belay run was absent on the second while the tumour was radiographically progressing, which is consistent with LP-WGS operating at or below its tumour-fraction floor. Belay Ascent has no published focal-CNV limit of detection or circulating-tumour-fraction QC, and its validation is vendor-authored and arm-level only; CSF cfDNA is a low-tumour-fraction substrate. The focal-CNV calls (RB1, MET, MDM4) are therefore low-reliability and are not used to drive decisions here; the pending SiMSen absolute MM/mL is the discriminating measure, and same-pipeline Belay QC reanalysis is recommended.

Belay call	Why low-reliability	Decision impact
chr13q14.2, RB1 deletion	Only DNA evidence is this run; otherwise mProbe-ND, Ignite pS780 (indirect), and an informal unsigned UCLA RB read (no signed IHC; RB status unproven)	RB1-loss not established; not a basis to contraindicate CDK4/6i; confirm via CSF RNA-seq
chr7q31, MET amplification	Prior DNA-confirmed label withdrawn	Hypothesis only (Ignite/IHC), pending CSF RNA-seq
chr1q32.1, MDM4 amplification	Call uninterpretable	Not interpretable; also not a target given mutant TP53 (Pre-therapy item 11)
chr7p EGFR / 9p-CDKN2A arm	Stochastic instability, same basis as above	EGFR not down-ranked and 9p not acted on from this run

Tracked-variant VAF kinetics · CSF · Aug 2025 → Apr 2026

Five tracked variants across Shanghai Decode (1200-gene, LOD 2%) and Belay (Summit 2.0, LOD 0.30%). The treatment-response decline (Feb→Mar) followed by an April rebound mirrors the independent SiMSen-Seq Day-7-of-CSI spike, the two monitoring programmes are cross-consistent.

Variant	Class	5 Aug 25 Decode	5 Feb 26 Decode/Belay	28 Feb 26 Decode	18 Mar 26	25 Mar 26	27 Mar 26	13 Apr 26 Belay
TP53 p.R273C	Truncal hotspot	1.89%	79.2%	20.13%	12.90%	9.62%	5.31%	24.60%
PDGFRA p.Y288C	Activating	—	28.5%	2.93%	1.95%	0.74%	1.04%	0.70%
H3-3A p.G35R	Diagnostic (WHO)	1.03%	25.4%	1.90%	5.78%	4.50%	2.33%	6.40%
ATRX p.E2270del	VUS / Class III	—	det.	18.64%	11.84%	12.40%	6.46%	23.50%
NF1 p.R156C	VUS, transient subclone	—	nd	17.70%	ND	ND	ND	ND
ctDNA (MTM/mL)	calc. from VAF	—	—	84.34	59.03	12.91	13.78	—

Interpretation, dilution-first (per Post-CSI MDT, 8 May 2026)

VAF is a ratio, not an absolute count. Six CSF-directed interventions sat between the 5 Feb and 13 Apr Belay draws (five intrathecal CAR-T doses plus Ad-TD-nsIL12 oncolytic virus); each releases non-tumour cfDNA into the compartment and inflates the denominator. The 12 April MRI showed marked LMD progression over the same interval. Apparent VAF declines must therefore be read against a 2–5× dilution-plausible band before any response inference. TP53 R273C (3.2×) and H3-3A G35R (4.0×) both fall within the dilution band and track each other, the signature of a shared denominator acting on a shared truncal clone, not confirmed response. PDGFRA Y288C (40.7×) is the only signal that exceeds dilution alone, but this should not be assumed to be on-target avapritinib activity, Y288C is resistant to the PDGFR TKIs tested by Ip 2018 (ER-trapped extracellular mutant; PMID 30389923; item 3); avapritinib was not tested (resistance plausible but unproven for this agent). A receding subclone ± residual denominator dilution is at least as parsimonious. The discriminating test is SiMSen-Seq absolute MM/mL on the current CSF, anchored to the 26 Feb baseline (TP53 R273C 2,095 MM/mL; total 11,592 MM/mL), pending.

chr7p / EGFR arm-call loss · subclones · neoantigens

The chr7p (EGFR) arm call was lost on Belay #2. This is not equivalent to EGFR clearance, three explanations remain live (on-target clearance, denominator dilution, subclonal/compartment shift) and cannot be separated on VAF data alone; EGFR should not be down-ranked without orthogonal CSF RNA-seq / SiMSen at the EGFR locus. NF1 R156C was a transient 28-Feb-only subclone (neoantigen source, now cleared). 28 Feb Shanghai Decode also: APOBR D710G (11.83%), NPAP1 D579N (4.96%), SELL G188V (2.01%, source of 2 HLA-A*02:06-restricted neoantigens, 3 neoantigens 28 Feb → 0 by 18 Mar). OncoIncytes blood ctDNA: no plasma signal, confirms CSF is the correct monitoring compartment.

Platform concordance

Biomarker	Belay 5 Feb	Decode 28 Feb–27 Mar	Belay 13 Apr	Concordance
MSI	Stable	MSS 1.0–3.6%	Stable	Concordant, MSS
TMB	Insufficient DNA	13.08 → 0.93 Muts/Mb	Low (measurable)	Concordant, Low
Aneuploidy / CIN	25 arm-level events	not assessed	14 arm-level events	Belay only, CIN reduction
9p loss (CDKN2A)	arm loss	not reported	arm loss	Belay only, relevant CDK4/6i & ICI
CDK12 P530A · AXIN2 Y550D	VUS	germline VUS Cl.3	VUS	Concordant, constitutional
HLA typing	het. all loci	het. all loci	—	Concordant, A*02:06 etc.

Platforms: Shanghai Decode 1200-gene (LOD 2%) · Belay Summit 2.0 520-gene + Ascent LP-WGS (LOD 0.30%) · OncoIncytes CTC+ctDNA. Source: ctDNA & NGS Monitoring workbook (CH Dashboard Drive). This series is parallel to and cross-consistent with the SiMSen-Seq personalised CSF kinetics in Axis 08.

Earliest CSF molecular timepoint · PredicineCARE 24 Jul 2025

The first CSF molecular sample, predating the 5 Feb 2026 Belay baseline by ~6 months. All VAFs are low (≤2.54%), consistent with low CSF tumour burden at that point (TP53 R273C was 1.89% here vs 79% by 5 Feb 2026). The detected set is enriched for DNA-damage-repair genes, a DNA-level correlate, in the CSF compartment, of the Axis 02 repair-deficiency picture.

Gene · alteration	VAF	Class	Axis	Note
RAD50 E723Gfs*5	2.54%	VUS (frameshift)	Ax02	Frameshift in the MRE11–RAD50–NBS1 DSB-repair complex. Highest-VAF CSF alteration at this timepoint.
TP53 R273C	1.89%	Truncal	Ax03	Same truncal hotspot as all other platforms; low VAF reflects early low burden.
H3F3A G35R	1.03%	Diagnostic	Ax06	Defining driver, present in every tumour cell.
BAP1 H141R	0.69%	VUS	Ax02	Deubiquitinase tumour suppressor; roles in HR / chromatin.
ATM H1951R	0.65%	VUS	Ax02	DSB checkpoint kinase, consistent with the DDR/checkpoint picture.
BRCA1 Q1556R	0.64%	VUS	Ax02	HR factor, adds to the repair-pathway lesion cluster.
GATA3 C375R · FAT1 D864G · MYD88 H156R	0.63–0.82%	VUS	—	Low-VAF VUS; uncertain significance.
MPL W515L	0.23%	VUS	—	Canonical myeloproliferative-neoplasm driver, most consistent with clonal haematopoiesis (CHIP), not tumour-derived. Flagged as a likely non-tumour signal.

Interpretation

The RAD50 / ATM / BRCA1 / BAP1 cluster is a DNA-level CSF correlate of the Axis 02 repair-deficiency picture and supports the niraparib + checkpoint-inhibitor rationale. All variants are VUS-class, directionally supportive, not independently confirmatory. MSI not detected; MMR genes (MLH1/MSH2/MSH6/PMS2) not detected on this panel.

Coverage note

NorthStar Select/Response (blood: probable-germline CDK12 P530A; methylation score stable), OncoIncytes (CSF ctDNA + CTC) and the Belay Summit 2.0 reports are already represented in the multi-platform monitoring table above and the Genomics tab; PredicineCARE is added here because it is the earliest CSF molecular timepoint and was not otherwise structured. Clinical-planning documents (Radiosensitiser strategy, Workplan, CH Work Doc) are operational, not molecular layers, and are not ingested into the data model.

Pre-therapy evidence review

Six evidence questions addressed before therapy selection: two reanalyses of existing data (mutational-process / HR-deficiency status; clonal architecture) and four targeted literature reviews (PDGFRA p.Y288C variant pharmacology; CSF-compartment delivery; ONC201 validity in G34; ATR inhibition in ALT / ATRX-loss). Drawn from the published literature and trial registries (2026). Each carries a graded confidence and, where existing data are insufficient, the specific orthogonal test required to resolve it.

Clonally safe to anchor therapy

TP53 R273C, truncal trunk (CSF + tissue)
H3.3 G34R, founding driver, truncal
ATRX-loss / ALT, truncal (IHC + SNV + fusion)
CDK6 amplification, truncal hard CNV

Subclonal / unstable, add-on only

PDGFRA Y288C, LMD-emergent, regressing
NF1 R156C, transient, cleared
ALK splice 7%, subclonal, non-canonical
EGFR (chr7p), indeterminate (arm-call lost)

Delivery / validity flags

PDGFRA Y288C, ER-trapped extracellular mutant resistant to the PDGFR TKIs tested by Ip 2018; avapritinib untested (futility plausible, unproven), not a well-supported backbone, resolve empirically
ONC201, no G34 evidence; demote from trunk
Ribociclib / elimusertib, fail CSF delivery gate
HRD scar, not measurable from data in hand; classical-HRD not supported by available proxies
Germline, DDR germline unconfirmed (enabling test)

1 · Mutational-process & HR-deficiency status, honest reanalysis of data in hand

The DDR-deficiency thesis (niraparib + topotecan, SMARCAL1 synthetic lethality) currently rests on protein non-detection (ND) of HR/NHEJ/MMR/NER effectors. ND is not equivalent to functional deficiency. A formal COSMIC SBS/ID signature deconvolution (SBS3 HRD, SBS11 temozolomide, SBS31/35 platinum, SBS6/15/20/26/44 MMR) and a Myriad-type genomic-scar HRD score (LOH + TAI + LST) cannot be computed from the data in hand, the Caris / Belay / Decode exports provide summary metrics, not the per-base mutation catalogue with trinucleotide context or allele-specific copy number those algorithms require. No signature percentages are reported here because none can be derived without the raw VCF/WGS; fabricating them would be unsafe.

What the scar proxies actually say

LOH burden LOW, Caris LOH "Low", 4% of segments. Classical BRCA-type HRD shows extensive LOH/LST. The low genomic-scar burden argues against a deep HRD scar.
MSS, declining TMB, MSI stable concordant across all platforms; CSF TMB 13.08 → 0.93 Muts/Mb (declining). Despite the reported MMR-protein loss, the genomic data show no active mutator / MSI-H phenotype, an MMR-d immunotherapy rationale is not supported.
CIN present but modest, 25 → 14 arm-level events; focal RB1 / MET / MDM4 / 9p, chromosomal instability without a high allelic-imbalance HRD signature.

Where the DDR rationale legitimately stands

The PARP/DDR case is supported by mechanistic and clinical-precedent evidence, germline-pattern CDK12 P530A (regulates BRCA1/RAD51 transcription), EME1 Q394* nonsense (MUS81–EME1 HJ resolution), ATRX-loss/ALT replication stress, Ignite CHK1/CHK2 95/99 pct, the exact-genotype Laemmerer 2025 niraparib+topotecan case, and the G34R DSB-repair-deficiency literature (Bčkaj 2022; Lo Cascio 2024; Lin 2024). It is not corroborated by an orthogonal HRD-scar/signature biomarker, and the available scar proxies are discordant with classical HRD.

Graded conclusion & exact tests to order

Confidence, functional DDR/replication-stress vulnerability: MODERATE (strong mechanistic + exact-genotype clinical precedent). Confidence, classical HRD-scar positivity: LOW (low LOH, MSS; no SBS3 available). The PARP/topotecan strategy remains defensible on functional/precedent grounds, but should be framed as a replication-stress / ALT vulnerability, not as biomarker-confirmed HRD. To close the gap, order: (i) tumour WGS or large panel with allele-specific CN → COSMIC SBS/ID signatures + HRD score (resolves SBS3 vs SBS11-TMZ vs SBS31/35-platinum vs MMR); (ii) RAD51-foci functional HR assay on FFPE block SSW-24-26425 (geminin-gated), most direct HR-deficiency readout, feasible on archival tissue; (iii) germline DDR panel (BRCA1/2, PALB2, RAD51C/D, Lynch/MMR; confirm CDK12 constitutional), currently "probable germline" is inferred from ~50% VAF only, and a true germline result changes PARP/platinum logic and family counselling; (iv) if immunotherapy considered: MMR IHC re-read + PCR-MSI on current CSF to reconcile MMR-protein loss against MSS / declining TMB.

2 · Clonal architecture & cancer-cell-fraction reanalysis

Reconstructed from variant allele frequencies already held: tissue WES (Caris, specimen SSW-24-26425, Oct 2024) and the serial CSF ctDNA series (Shanghai Decode 1200-gene LOD 2%; Belay Summit 2.0 LOD 0.30%), Aug 2025 → Apr 2026, read under the established dilution-first framework. Decision question: which targets are truncal (safe to anchor therapy) versus subclonal / compartment-emergent (escape risk).

Tissue purity caveat

H3F3A G34R sits at 17% VAF in tissue. For a truncal heterozygous driver with no local CNV, VAF ≈ purity/2, implying tumour cellularity ≈34%, consistent with an infiltrative, normal-brain-diluted specimen. TP53 R273C at 45% cannot be reconciled with that purity under simple heterozygosity (expected ≈17%), indicating regional purity heterogeneity and/or 17p LOH. Absolute tissue CCF is therefore unreliable for an infiltrative specimen; the SiMSen-Seq absolute CSF burden (anchored 26 Feb: TP53 R273C 2,095 MM/mL; total 11,592 MM/mL) is the correct clonal-burden metric for this compartment disease.

Variant	Class	5 Aug 25	5 Feb 26	28 Feb	18 Mar	25 Mar	27 Mar	13 Apr	Clonal verdict
TP53 p.R273C	Truncal hotspot	1.89%	79.20%	20.13%	12.90%	9.62%	5.31%	24.60%	TRUNCAL trunk. Highest VAF at every informative point; SiMSen-anchored. Safe to anchor (p53 axis: R273C + MDM4-amp).
H3-3A p.G35R (G34R)	Diagnostic driver	1.03%	25.40%	1.90%	5.78%	4.50%	2.33%	6.40%	TRUNCAL. Co-varies with TP53 at a stable ~0.26–0.32 ratio → shared clone + shared denominator, not subclonality.
PDGFRA p.Y288C	Activating	—	28.50%	2.93%	1.95%	0.74%	1.04%	0.70%	SUBCLONAL, LMD-emergent, regressing. Absent 2024 tissue & Aug 2025; PDGFRA/TP53 ratio 0.36→0.03 (×13, exceeds the 2–5× dilution band). Y288C is resistant to the PDGFR TKIs tested by Ip 2018 (ER-trapped; item 3; avapritinib untested) — the decline is at least as likely a receding subclone ± dilution as on-target avapritinib effect. Not a well-supported backbone target.
ATRX p.E2270del	VUS / det.	det.	18.64%	11.84%	12.40%	6.46%	—	23.50%	TRUNCAL. Tracks the trunk; corroborates ATRX-loss (IHC + ATRX:MAGT1 fusion). ALT/DDR strategy targets a truncal vulnerability.
NF1 p.R156C	Transient subclone	—	nd	17.70%	ND	ND	ND	ND	TRANSIENT. 28-Feb-only; neoantigen source, now cleared. Not a stable target.

Decision implication

Therapy should be anchored on the truncal set, p53 axis (TP53 R273C + MDM4-amp), H3.3 G34R lineage biology, ATRX-loss/ALT replication-stress vulnerability, and the truncal CDK6 amplification. These are present in (near-)all tumour cells, so on-target activity cannot be escaped by a pre-existing major clone.

PDGFRA Y288C, the key caveat

Y288C is not truncal (recurrence/LMD-emergent; CSF fraction down ~40× vs baseline, ~13× vs trunk) and is resistant to the PDGFR TKIs tested by Ip 2018 (ER-trapped extracellular mutant — item 3; avapritinib not tested, futility plausible but unproven). A PDGFRA-directed TKI is therefore not a well-supported backbone; the better-grounded targets for this signal are downstream PI3K/mTOR and MEK. Avapritinib activity is best resolved empirically with on-treatment SiMSen absolute MM/mL during any avapritinib exposure (reconciled against the drug diary) before any PDGFRA-led decision.

3 · PDGFRA p.Y288C, variant-specific pharmacology

PDGFRA p.Y288C is an extracellular-domain mutant. A naïve reading would extend the D842V→avapritinib (type-I) logic to it; a variant-specific reanalysis shows that does not transfer, while also calibrating what is and is not proven for avapritinib specifically.

Where residue 288 sits

Y288 is extracellular, in the third Ig-like C2-type domain (D3) of PDGFRA (UniProt P16234: signal peptide 1–23; extracellular ~24–524 = five Ig-like domains, D3 ~225–308 contains residue 288; TM ~525–548; kinase ~593–954, with the avapritinib-sensitive D842V in the exon-18 activation loop). It is not a kinase- or juxtamembrane-domain mutation, so its drug-response biology differs fundamentally from D842V.

Direct (not extrapolated) functional data on Y288C

Ip, Mills et al. Nat Commun 2018 (PMID 30389923, 10.1038/s41467-018-06949-w) characterised Y288C by name: high-mannose glycosylated, ER-trapped, constitutively dimerised and phosphorylated ligand-independently, driving constitutive AKT/ERK/STAT3. Y288C was resistant to the PDGFR inhibitors tested (older / conformation-dependent agents, imatinib, dasatinib, crizotinib class) and sensitive to PI3K/mTOR and MEK inhibition. Title verbatim: "Neomorphic PDGFRA extracellular domain driver mutations are resistant to PDGFRA targeted therapies". Important calibration: avapritinib was not among the agents tested in Ip 2018, and the authors explicitly noted it as an untested inhibitor that should be evaluated against these mutants. Avapritinib resistance for Y288C is therefore mechanistically inferred (ER-trapped, oncogenicity ligand- and kinase-conformation-independent, so an ATP-site type-I inhibitor is unlikely to rescue), plausible but not directly demonstrated for avapritinib. This is why the recommendation is data-gated equipoise resolved empirically, whenever avapritinib is given, the SiMSen on-treatment burden is the arbiter (exposure timing to be reconciled against the drug diary), not a categorical exclusion.

Why the D842V→avapritinib logic does not transfer

Avapritinib’s D842V activity is a type-I inhibitor binding the kinase active conformation (Rizzo 2021, 10.3390/cancers13040705; Grunewald/Bauer 2021, 10.1158/2159-8290.CD-20-0487). Y288C oncogenicity is not a druggable kinase-conformation problem — it is ER-retention/aberrant-glycosylation driven. Glioma-specific corroboration: Mayr/Filbin/Koschmann, Cancer Cell 2025 (PMID 40086436, 10.1016/j.ccell.2025.02.018) — avapritinib non-responders harboured extracellular-domain mutants, whereas PDGFRA-amplified tumours responded; they explicitly cite Y288C as ER-trapped/PDGFR-inhibitor-resistant. Funato/Tabar (10.1016/j.cell.2020.11.012) explain the recurrence-selected emergence but report no Y288C-specific drug sensitivity.

CNS delivery is not the limiting factor

Avapritinib does cross the BBB with measured human CNS exposure and intracranial responses (Mayr 2025) — the only PDGFRA inhibitor with glioma CNS-penetrance evidence. Dasatinib historically fails on CNS PK/tolerability. So for Y288C the limitation is the variant class, not delivery: a CNS-penetrant TKI still cannot rescue an ER-trapped, kinase-conformation-independent mutant.

Graded conclusion & action

Confidence that a Y288C-directed PDGFR TKI has a strong rationale: LOW. Ip 2018 shows resistance across the (older, conformation-dependent) PDGFR TKIs it tested; the avapritinib-for-D842V type-I paradigm does not transfer to an ER-trapped extracellular mutant; the glioma series places extracellular mutants among non-responders. However, avapritinib itself was not tested by Ip 2018 and the authors explicitly nominated it for evaluation, so avapritinib futility for Y288C is mechanistically plausible but unproven, not categorical. Combined with the clonal reanalysis (item 2 — Y288C is subclonal, LMD-emergent, regressing, not truncal): (1) a Y288C-directed PDGFR-TKI is not a well-supported backbone; (2) the better-grounded targets for this signal are downstream PI3K/mTOR and MEK (direct Ip 2018 evidence); (3) avapritinib’s effect on the Y288C clone is directly testable, resolve empirically with on-treatment SiMSen absolute MM/mL during any avapritinib exposure rather than by assumption. The apparent CSF PDGFRA-VAF decline should not be assumed to be on-target avapritinib activity; a receding subclone ± denominator dilution is at least as parsimonious, and the empirical burden test discriminates them.

4 · CSF / leptomeningeal delivery gate, candidate agents

For leptomeningeal disease an agent that does not reach the CSF compartment at active concentrations fails regardless of target match. Note: most "brain-penetration" data are parenchymal tumour-tissue measurements; parenchymal Kp does not guarantee leptomeningeal/CSF coverage (the LM compartment behaves as a relative sanctuary). As established in the published literature (with Neuro-Oncology / ASCO and trial sources where journal data on CSF delivery were sparse).

Agent	CSF / CNS exposure evidence	Efflux	Verdict	Preferred route
Topotecan	IV: CSF:plasma ~0.29–0.42; AUC ratio ~0.11–0.19 (PMID 8529278; 7923128; 16461424). Established intrathecal formulation, low grade≥3 toxicity (doi 10.1177/10781552261422155).	BCRP + P-gp	GOOD via IT / MARGINAL IV	IT / Ommaya
Avapritinib	Brain + spinal-cord penetration; CSF:plasma rises over time, tumour ~4µM, CSF » dasatinib; clinical CNS responses (PMID 40086436).	—	GOOD (measured CSF + cord)	Systemic
Niraparib	Ivy LC-MS/MS: measurable brain + CSF; unbound 226.7 nM in non-enhancing GBM, 100% of n=44 met PK threshold; NCT05076513 (doi 10.1016/j.jpba.2024.116150).	P-gp (high passive perm. overcomes)	GOOD parenchymal / MARGINAL pure-LM	Systemic
ONC201 (dordaviprone)	BBB-penetrant by design; pediatric DIPG PK + CSF target-engagement (2-HG) NCT03416530 (doi 10.1093/neuonc/noae001); quantitative CSF:plasma not published.	—	GOOD parenchymal / UNKNOWN CSF	Systemic (oral)
Selinexor	Resected-tumour median 105.4 nmol/L; RANO responses, PFS6 17% rGBM NCT01986348 (doi 10.1158/1078-0432.CCR-21-2225); no human CSF ratio.	—	MARGINAL parenchymal / UNKNOWN CSF	Systemic (unproven)
Abemaciclib	PBPK CNS model: outperforms ribociclib but limited; class poor BBB permeability (doi 10.1002/cpt.3505).	P-gp/BCRP	MARGINAL CSF	Systemic (weak)
Ceralasertib (ATRi)	Preclinical effective brain penetration (doi 10.1186/s13014-018-1020-3); no human CSF data.	—	MARGINAL (preclinical only)	Systemic (unproven)
Ribociclib	Documented poor BBB penetration; dedicated brain-penetrant analog programmes (doi 10.1016/j.bmcl.2019.06.021).	P-gp/BCRP	POOR	Exclude
Elimusertib (ATRi)	CNS distribution P-gp-limited, high tissue binding → low unbound brain drug; "would not maintain adequate CNS exposure".	P-gp	POOR	Exclude (systemic)

Bottom line

For this leptomeningeal patient, intrathecal / intra-Ommaya topotecan is the only candidate with established CSF-compartment delivery; avapritinib is the strongest systemic agent with directly measured CSF + spinal-cord exposure (target match contingent on the Y288C clone, see item 2–3); niraparib has the best systemic parenchymal dossier but pure-LM CSF coverage is unquantified. Ribociclib and elimusertib fail the delivery gate regardless of target match, if a CDK4/6 arm is pursued, abemaciclib is the less-weak choice; if ATR inhibition is pursued, CNS delivery is the principal uncertainty.

5 · ONC201 (dordaviprone) validity in H3 G34R

ONC201’s efficacy signal is defined by H3 K27M, whereas this tumour is H3 G34R, its validity in this genotype requires explicit appraisal. As established in the published literature and the FDA/Jazz approval record.

What the evidence shows

Mechanism is genotype-agnostic, mitochondrial ClpP agonism + DRD2 antagonism; not K27M-dependent (JCO 2024 10.1200/JCO.23.01134; Cancer Res 2023 10.1158/0008-5472.CAN-23-0186). So ONC201 is not biologically impossible in G34R.
All pivotal efficacy is H3 K27M, registrational n=50, all recurrent K27M DMG, ORR ~20%; FDA accelerated approval (Modeyso, Aug 2025) and Phase 3 ACTION (NCT05580562) are K27M-restricted. The "cerebral" expansion (Odia 2024) was still K27M-mutant.
Zero direct G34 / ALT data, exhaustive search returned no cell-line, PDX, ClpP/DRD2-expression or clinical data for ONC201 in H3 G34 or ATRX-loss/ALT glioma. Indirect DRD2/lineage rationale is hypothesis-generating only and ClpP (not DRD2) is now considered the primary effector.

Graded conclusion

NOT JUSTIFIED as trunk therapy. ONC201’s entire efficacy signal, approval, and patient-selection biomarker are defined by H3 K27M; this patient is H3.3 G34R, a distinct entity with no ONC201 data of any kind. Keeping it in a trunk is precisely the K27M-by-analogy anchoring to avoid. Action: remove from the trunk; retain only as an explicitly-labelled exploratory / clinical-trial option. Single deciding reason: biomarker-defined efficacy population does not include this genotype, and no preclinical bridge exists.

6 · ATR inhibition in ALT / ATRX-loss glioma

ATRX-loss → ALT confers acute ATR-dependence of telomere maintenance. The plan currently leaves ATR inhibition implicit. As established in the published literature and trial registries.

Mechanistic & glioma-specific support

ALT → ATR-dependence, Flynn/Zou Science 2015 (10.1126/science.1257216): ATRX-loss makes ALT telomere maintenance acutely ATRi-sensitive (glioma lines included). Context-dependent (Deeg 2016; Gonçalves 2020, tracks short telomeres).
Glioma/ATRX-specific, Garbarino/Bindra Transl Oncol 2021 (10.1016/j.tranon.2021.101147): isogenic glioma, ATRX-loss → replication stress + constitutive ATR activation + PARPi sensitivity + synergistic ATR+PARP killing. Yuan/Heaphy 2022 (10.3390/cancers14123015): ATRX-loss/ALT sensitises U251 to ATRi. No G34-specific study — inference from ATRX/ALT biology.
Caution, Koneru/Reynolds 2021 (10.1126/scitranslmed.abd5750): AZD6738 did not reverse ALT chemoresistance in neuroblastoma (ATM-driven there).

Agents, combination logic & positioning

Elimusertib (BAY1895344), best CNS candidate (dedicated CNS-distribution data; pediatric NCT05071209). Camonsertib (RP-3500) has the first clinical ALT+ response (10.1038/s41698-025-01025-1) but no CNS data; ceralasertib limited CNS; berzosertib not oral.
ATRi is upstream-redundant with CHK1/WEE1 (same ATR–CHK1 axis as the Ignite CHK1/CHK2 95/99 pct finding) — pick one node, do not stack. ATRi is non-redundant with PARPi/topotecan and converges with the SMARCAL1 replication-fork concept.
Safety, niraparib + topotecan is already heavily myelosuppressive; adding ATRi risks intolerable overlapping thrombocytopenia/neutropenia. Triplet not established.

Graded conclusion

Justification: MODERATE–HIGH (strong mechanistic + glioma-specific preclinical incl. ATR+PARP synergy; low direct G34/clinical evidence; CNS delivery the principal uncertainty — see item 4). Position as a distinct, sequential / alternative arm (elimusertib), not added to the niraparib+topotecan backbone, and let it replace rather than duplicate any CHK1/WEE1 arm. Reserve for niraparib/topotecan failure or progression, optionally as ATRi + low-dose PARPi.

7 · Short-read vs long-read genome-wide concordance, the decisive reanalysis

Caris short-read TPM Z vs UCLA long-read CPM Z, computed genome-wide on 13,249 shared genes. Pearson r = 0.383, weak. Of 663 genes Caris calls high-outliers (Z > +2), 273 (41%) are NOT corroborated by long-read (ΔZ ≥ 2). The genes used elsewhere to argue PDGFRA over-expression and a "downstream independently over-active" combination are precisely the discordant ones; CDK6 is the one driver that is cross-platform concordant.

Gene	Caris Z	LR Z	ΔZ	LR logFC vs GTEx	Verdict
CDK6	+1.98	+3.08	−1.10	+6.15	CONCORDANT, robust
CCND2	+3.76	+2.48	+1.29	+2.50	concordant
PDGFRA	+3.87	+0.80	+3.07	−1.72	DISCORDANT (SR-high; LR below normal brain)
PDGFRB	+3.03	−0.19	+3.22	−1.37	DISCORDANT
KIT	+2.30	−0.45	+2.74	−2.07	DISCORDANT
PIK3CA	+2.58	−0.92	+3.50	−2.28	DISCORDANT
AKT3	+4.71	−0.67	+5.38	−4.22	DISCORDANT
MTOR	+2.73	−0.12	+2.86	+0.17	DISCORDANT
MAP2K1	+2.97	+0.11	+2.86	−6.84	DISCORDANT
MAPK1	+3.55	+0.20	+3.35	−5.54	DISCORDANT
EGFR	+3.24	−0.05	+3.28	+0.07	DISCORDANT
RB1	+2.45	+0.93	+1.52	−1.38	mild

Consequence

The Caris short-read transcriptome is only weakly reliable here, so any single-platform short-read over-expression call is suspect. (1) The PDGFRA over-expression rationale collapses, long-read places PDGFRA/PDGFRB/KIT below normal brain; combined with item 3 (Y288C resistant to the PDGFR TKIs Ip 2018 tested, avapritinib untested, futility plausible but unproven, and subclonal) this places avapritinib in data-gated equipoise. (2) The "vertical co-blockade because PI3K/AKT/MEK are independently over-active" argument loses its transcriptomic leg (AKT3/MAPK1/MAP2K1 long-read 4–7 logFC below brain); pursue only if Ignite RPPA phospho-activation supports it, never on mRNA. (3) CDK6 over-expression is corroborated by both platforms and is +6.15 logFC vs GTEx, the single most robust driver signal, strengthening CDK4/6 inhibition as the primary arm.

8 · Belay Ascent LP-WGS reliability appraisal

The truncal chr7p/EGFR arm gain present on Belay #1 was absent on Belay #2 while the tumour was radiographically progressing. Drawn from the published literature and vendor-disclosed assay specifications.

Why the EGFR disappearance is an artifact signature

A near-universal early GBM driver cannot vanish during progression. Present-then-absent arm calls in a progressing tumour are the textbook signature of an LP-WGS assay at/below its circulating-tumour-fraction floor, stochastic call instability mimicking subclonal clearance. The whole run inherits unquantified instability; true clearance would require concordant SNV-VAF + copy-burden + radiographic falls (the opposite occurred here).

Assay limits

No published focal-CNV LOD, no ctF floor, no per-sample ctF/MAD/ploidy QC. Only peer-reviewed validation (Nie 2025, JMD, PMID 40280408) is vendor-authored and validates only an arm-level aneuploidy LOD.
LP-WGS copy-number unreliable <~3% tumour fraction (ichorCNA/Adalsteinsson 2017, PMID 29109393). CSF cfDNA is worst-case (CNAs in only ~20% of CSF, Pagès/Adalsteinsson, Neuro Oncol 2022, PMID 34984433).

Verdict, LOW → UNINTERPRETABLE for the RB1/MET/MDM4 focal calls

Do not use Belay Ascent focal CNVs to drive de-escalation or target selection. RB1-del → discount (removes the main argument against CDK4/6i). MET-amp → downgrade to hypothesis (Ignite/IHC-supported only). MDM4-amp → doubly off (uninterpretable + non-target with mutant p53, item 11). To rehabilitate: matched same-pipeline reanalysis with per-sample ctF/MAD/ploidy QC + orthogonal confirmation. Converges with the treating team's Post-CSI MDT (Belay QC flagged CRITICAL).

9 · Treatment history & response reconciliation

The patient has progressed through ~12 prior lines, including prior avapritinib and abemaciclib exposure; current and recent therapy must be reconciled against the drug diary by the treating team (this document does not assume an active regimen).

Failed / active prior lines

Surgery ×2, chemoRT + TMZ, CCNU ×5, DCVax, Optune, ImmunityBio NK, SRS, NK-cell, all progressed.
ONC201/206 since Mar 2025 (escalated Jan 2026), progressed (empirical failure, not just off-G34-biomarker, item 5 strengthened).
CAR-T ×4 (EGFR/B7H3, ± GD2), radiographic progression through every cycle; G4 lymphopenia (ALC 0.21) precluded T-cell expansion → all T-cell immunotherapy (incl. PRAME-TCR-T) down-graded pending lymphocyte recovery.
Avapritinib 150 mg + abemaciclib 75 mg + Ad-TD-nsIL12 OV active; CSI proton 13 Apr→8 May.

Response & data-integrity flags

12 Apr MRI = worst to date; 30 Apr/1 May = first non-progression since Nov 2025, MIXED, not a formal RANO PR (dashboard "May 4 PR" corrected).
ctDNA draw dates in the multi-platform tab do not match source documents, flag for QC.
CAR-T count 4 vs 5 and avapritinib stop-date are unreconciled in source docs.
Treating-team Post-CSI MDT independently converges with this plan (RETAIN avapritinib pending data, GATE abemaciclib on CSF RNA-seq RB1, PARPi 4–8 wk post-CSI count-gated, MDM4 not actionable).

10 · Absolute MM/mL clonality (SiMSen), confirms dilution-first

VAF ratios are confounded by the unstable cfDNA denominator; SiMSen absolute MM/mL is less affected by cfDNA dilution than VAF. Anchored timepoints: 26 Feb baseline & 20 Apr (CSI Day 7).

Variant	26 Feb MM/mL	20 Apr (CSI D7)	Fold	Status
Total cfDNA	11,592	48,989	+4.23×	acute CSI release flood
TP53 R273C (trunk)	2,095	10,697	+5.11×	trunk burden ROSE, no prior decline = response
MTIF3 L157V	106	1,529	+14.4×	watch
CCR3 I223V	458	2,796	+6.11×	watch
MICAL2 / ALK / MIR6817	—	—	<1×	receding sub-clones

Reading

Trunk absolute burden rose 5× into CSI Day 7 (acute radiation release), no earlier Belay VAF decline is confirmed response. PDGFRA Y288C is not in the SiMSen personalised panel, so its clonal status cannot be cross-checked in absolute terms, the pending SiMSen Day-16/29/12-May MM/mL (PDGFRA-spiked if possible) is the single sound discriminator and the gate for the avapritinib continue/stop decision.

11 · TP53 R273C + MDM4-amp, a checkpoint node, not a p53/MDM4 target

Whether the p53 axis is druggable.

Why p53/MDM4 is NOT the target

MDM2/MDM4 inhibitors (incl. ALRN-6924) require wild-type p53; truncal R273C (high CSF VAF, likely WT-allele loss) makes this inapplicable. MDM4-amp = p53-axis-inactivating context, not a target.
R273C is a DNA-contact mutant, outside the rezatapopt (Y220C-only) and ATO/ZMC structural-mutant sweet spot; APR-246 effect is really glutathione-depletion/ferroptosis, not R273C reactivation.

The actionable consequence

p53-deficiency → loss of the G1/S checkpoint → dependence on the G2/M (WEE1/CHK1) checkpoint, which connects directly to the Ignite RPPA CHK2 99 / CHK1 95 pct activation and to radiosensitisation under CSI. The genuine lever is WEE1/CHK1 checkpoint exploitation, not p53 reactivation. A statin (GOF/mevalonate, contact-mutant-selective) is a low-confidence adjunct only.

Graded conclusion

Druggable only via downstream checkpoint exploitation (moderate-strong). Reframe the "p53 + ferroptosis" axis: the MDM4/p53-reactivation reading is withdrawn; the actionable node is a single brain-penetrant WEE1 inhibitor leveraging p53-loss + the Ignite CHK signal + CSI radiosensitisation. ATRi (item 6), CHK1i and WEE1i are mutually redundant, pick ONE checkpoint node.

Therapy recommendations

Ranked, colour-coded therapy options, available now trial / near-term stop hold, each with a graded confidence, the deciding reason, a suggested 6-week combination, and the gating test. Provisional, for multidisciplinary review, not a fixed protocol; active therapy to be reconciled against the drug diary.

Ranked recommendations

1 · CDK4/6 inhibition, abemaciclibAVAILABLE · primary · MOD–HIGH

The only cross-platform-concordant truncal driver (long-read +6.15 logFC vs GTEx; CDK6:VIPR2 amplicon-rearrangement; abemaciclib > ribociclib in the cell screen and on CNS PK). RB1-loss concern is unresolved (Belay LP-WGS uninterpretable; no formal RB IHC, RB status unproven). Gate: CSF RNA-seq RB1/CDK6 + RB1 IHC.

Suggested 6-week combination: Abemaciclib + everolimus (the convergent PI3K/mTOR partner, covers CDK4/6-escape and the PDGFRA-Y288C downstream signal) ± avapritinib (low-toxicity add-on, see 3).

2 · Replication-stress / ALT, niraparib + intrathecal topotecanAVAILABLE · MODERATE

Truncal ATRX-loss/ALT; exact-genotype Laemmerer-2025 precedent. Niraparib has the best PARPi CNS PK; intrathecal/Ommaya topotecan is the CSF-delivery-validated partner for the leptomeningeal compartment. Timing: 4–8 wk post-CSI, count-gated (PLT>100/ANC>1.5/Hgb>10).

Suggested 6-week combination: Niraparib (systemic) + topotecan intra-Ommaya/IT, as a sequenced block, not stacked with everolimus+MEK (overlapping myelotoxicity).

3 · Avapritinib, retain & adjudicate (favourable tolerability)AVAILABLE · EQUIPOISE

Ip 2018 did not test avapritinib (resistance unproven in either direction); it is CNS-penetrant with measured intracranial exposure and a clean cell-line toxicity profile, a genuinely low-risk addition while the on-treatment SiMSen absolute MM/mL adjudicates. Not a backbone, but the low downside justifies inclusion in the available combination rather than holding it back.

Suggested 6-week combination: Add avapritinib to abemaciclib + everolimus (minimal added toxicity, full CNS coverage); continue/stop on the SiMSen burden read at the 6-week gate.

4 · Single checkpoint node, WEE1i (Debio 0123)TRIAL/NEAR · MOD-strong (mech)

p53-loss G1/S abrogation + Ignite CHK2 99 / CHK1 95 pct + CSI radiosensitisation converge here. Purpose-built brain-penetrant (adavosertib excluded, P-gp/BCRP). WEE1i, ATRi and CHK1i are mutually redundant, one only, and not concurrent with the PARP backbone (myelotoxicity).

Suggested 6-week combination: Debio 0123 ± low-dose alkylator/RT (trial NCT05765812); replaces, does not stack with, recommendation 2.

5 · Selinexor ± RT, priority experimentalTRIAL · HYPOTHESIS+

Biologically anchored: cross-platform-concordant NF-κB/IκBα (NFKBIA) context, selinexor’s core mechanism, plus potent, deep-blue cytotoxicity across all four screen lines, reproducible recurrent-HGG activity and measured CNS penetration. Not the fusion (passenger) or XPO1-expression (short-read-only).

Suggested 6-week combination: Selinexor ± RT (trial NCT05099003); reasonable as an early intensification partner or switch, ranked above other experimental options.

6 · MEK / RAF–FAK arm, avutometinib + defactinibTRIAL · HYPOTHESIS

Second downstream-of-Y288C node (Ip 2018: Y288C is MEK-sensitive) and pre-empts MAPK/RTK-switch bypass of the spine. Trial explicitly includes DHG H3 G34.

Suggested 6-week combination: Spine (abemaciclib+everolimus) + avutometinib/defactinib (NCT06630260) as a stable/mixed-response intensification; watch overlapping toxicity.

7 · ONC201 / ONC206, discontinueSTOP · stop

Empirically failed in this patient (on it Mar 2025 → escalated Jan 2026 → progressed), with no H3 G34 efficacy data of any kind.

Suggested 6-week combination: Discontinue; do not carry into any block.

8 · Temozolomide, excludeSTOP · stop

MGMT-methylated + MMR-protein loss + prior TMZ + rising TMB: MMR-loss converts TMZ from cytotoxic to a hypermutagen (SBS11; not checkpoint-responsive). MGMT-methylation alone is insufficient. Decisive test: MMR IHC + MSI + SBS11 signature on current tissue. CCNU/BCNU are a distinct crosslinking mechanism (different risk, not "safe").

Suggested 6-week combination: Do not re-challenge TMZ; exclude from all blocks pending the SBS11/MMR result.

9 · T-cell immunotherapy / PRAME-TCR-T, holdHOLD · defer

CAR-T ×4 progressed through every cycle with grade-4 lymphopenia (ALC 0.21) precluding T-cell expansion.

Suggested 6-week combination: Hold all T-cell approaches pending lymphocyte recovery; revisit only if counts recover.

CNS-delivery resolution, the load-bearing PK check

For leptomeningeal disease, CSF-compartment delivery can matter more than systemic BBB penetration. Each node is resolved to a named agent and a delivery route.

Node	Agent	Systemic CNS (BBB)	Intrathecal / intra-Ommaya	Verdict
CDK4/6	Abemaciclib	Modest, best of the CDK4/6 class (ribociclib/palbociclib poorer)	No IT data	Systemic; gate on RB1
PI3K/mTOR	Everolimus	Limited but G34-trial-supported (NCT05843253)	No IT data	Systemic partner
Replication-stress	Niraparib + topotecan	Niraparib: best PARPi CNS PK (Ivy LC-MS/MS)	Topotecan IT established (intraventricular MTD 0.4 mg; arachnoiditis is the DLT)	IT-topotecan delivers the LMD compartment
Checkpoint	Debio 0123	Purpose-built brain-penetrant; adavosertib P-gp/BCRP-restricted (excluded)	No IT data (no WEE1i IT)	Systemic; LM coverage extrapolated
PDGFRA	Avapritinib	Measured intracranial exposure / CNS responses	No IT data	Systemic, CNS-validated
XPO1	Selinexor	Resected-tumour exposure (~105 nmol/L); rGBM activity	No IT data	Systemic only
LMD cytotoxic adjunct	Pemetrexed (IT)	Systemic CNS limited	Credible IT option, phase 1/2 NSCLC-LM: 50 mg IT + dexamethasone, ~80% RANO response (Fan 2021/2024); arachnoiditis mitigated by dexamethasone	Deliverable IT for LMD, discuss as a CSF-directed adjunct
Established IT comparators	MTX · cytarabine/liposomal-cytarabine · thiotepa · etoposide	—	Established IT agents	Standard IT backbone options for LMD control
NOT IT-deliverable	Anthracyclines · bortezomib · irinotecan/SN-38 · selinexor	—	No human IT data / neurotoxic	Do not pursue intrathecally

Principle

Favourable for IT: small/water-soluble or sustained-release formulation, low arachnoidal neurotoxicity, predictable CSF clearance. Unfavourable: intrinsic neurotoxicity (anthracyclines, bortezomib), toxic excipients (Cremophor-paclitaxel), or simply absent human IT data (irinotecan, selinexor). Combined IT + brain-penetrant systemic therapy raises leukoencephalopathy risk.

Gating tests (precede / accompany)

CSF RNA-seq RB1/CDK6 expression · RB1 IHC (2024 block) · pending SiMSen absolute MM/mL (PDGFRA-Y288C- and CCNE1-spiked) · RAD51-foci HR assay (2024 block) · germline DDR panel · MMR IHC + MSI + SBS11 mutational-signature on current tissue (decisive for TMZ exclusion) · matched same-pipeline Belay reanalysis with per-sample ctF/MAD/ploidy QC before any Belay CNV is acted on.

Bottom line

Lead with the available tier: abemaciclib + everolimus (CDK4/6 truncal + the convergent PI3K/mTOR partner that also covers PDGFRA-Y288C), with avapritinib added as a low-risk equipoise component (well-tolerated, CNS-penetrant, adjudicated on SiMSen) and the post-CSI, count-gated niraparib + intrathecal-topotecan replication-stress block sequenced next. Trial/near-term: a single brain-penetrant checkpoint node (Debio 0123) and selinexor ± RT (biologically anchored, prioritised). Stop ONC201/206 and exclude temozolomide (hypermutation risk). Hold T-cell therapy. CNS delivery is the load-bearing constraint, intrathecal topotecan (and, for LMD adjunct, intrathecal pemetrexed) are the CSF-compartment-deliverable options.

6-week response-adaptive blocks

Therapy is delivered in discrete 6-week blocks rather than a fixed long-horizon schedule. Day 0 of every block: CSF SiMSen absolute ctDNA (absolute MM/mL, not VAF), amino-acid PET (FET / F-DOPA, TBR_max), and MRI (RANO, brain + whole-spine for LMD). Week 6: repeat the same triad, generate a composite response call, and branch accordingly, a rolling, response-adaptive loop. Provisional, single-patient.

1 · Block engine, monitoring & response definition

Modality	Reads	Primary metric
CSF SiMSen ctDNA	Absolute molecular burden / clonal kinetics	Trunk TP53 R273C MM/mL + tracked variants; add PDGFRA Y288C & CCNE1 (see item 4–5)
Amino-acid PET	Metabolically active tumour vs treatment effect	TBR_max change
MRI brain + spine	Anatomic burden, LMD, mass effect	RANO (bidimensional + LMD nodularity)

Composite call at each 6-wk gate

Responding, ctDNA MM/mL ↓ + PET TBR ↓ + MRI stable/↓ → roll same block forward.
Stable / mixed, discordant or flat → intensify (add the next escape-aware partner).
Progression, ctDNA rise (pre-radiographic) and/or new lesion / PET ↑ → switch block.

Two escape biomarkers to add to the SiMSen panel now

PDGFRA p.Y288C, currently absent from the personalised panel (a monitoring gap); it is the PDGFRA-clone escape marker.
CCNE1 / cyclin E1 (± PIK3CA, TP53), rising ctDNA predicts CDK4/6i escape and precedes radiographic PD by months → the early-switch trigger.

2 · The spine, why this combination (convergence)

CDK6-inhibitor escape and the PDGFRA-Y288C axis both funnel into PI3K/mTOR (± MEK). A brain-penetrant CDK6 backbone (abemaciclib) + everolimus (PI3K/mTOR) therefore does three things at once: addresses the CDK6 lineage dependency, blocks the dominant CDK4/6i bypass, and is the better-grounded PDGFRA-Y288C-axis intervention (Ip 2018: Y288C resistant to the PDGFR TKIs tested, avapritinib untested; sensitive to PI3K/mTOR & MEK). It maps to a real G34-inclusive trial (NCT05843253, ribociclib/abemaciclib + everolimus). The MEK arm (avutometinib + defactinib platform, NCT06630260, explicitly includes DHG H3 G34) is the intensification / alternative downstream node.

3 · Block decision tree

Block 1 · Wk 0–6 · SPINE

Abemaciclib + everolimus. Stop avapritinib unless tissue/ctDNA shows a concurrent conformation-competent PDGFRA amplification (then avapritinib as add-on only, not for Y288C alone). Baseline triad Day 0; gate at Wk 6.

Wk-6 gate → branch

Responding → Block 2-CONTINUE
Stable/mixed → Block 2-INTENSIFY
Progression → Block 3-SWITCH

Block 2-CONTINUE · Wk 6–12

Same spine (abemaciclib + everolimus). Repeat triad Wk 12; same 3-way branch.

Block 2-INTENSIFY · Wk 6–12

Spine + MEK (avutometinib platform, NCT06630260, DHG-G34-inclusive), second Y288C-defensible downstream node + pre-empts MAPK/RTK-switch bypass. Watch overlapping toxicity; do not also add the myelosuppressive backbone here.

Block 3-SWITCH · on PD

Replication-stress / checkpoint block: niraparib + IT/Ommaya topotecan (count-gated, ≥4–8 wk post-CSI; PLT >100 / ANC >1.5 / Hgb >10) OR a single brain-penetrant WEE1i (Debio 0123) if the Ignite checkpoint signal dominates OR selinexor ± RT (priority experimental, NF-κB/IκBα-anchored; trial NCT05099003), which may also be considered as a Block-2 intensification rather than held to last line. Not stacked with everolimus + MEK (myelotoxicity). Repeat triad at 6 wk.

Rolling loop

Every block ends with the same triad and the same 3-way branch, adaptive, response-driven, indefinitely re-enterable. No agent runs unmonitored beyond 6 weeks.

4 · CDK6, resistance, plasticity & how the combination exploits it

Why CDK4/6i fails, escape biology

G34 DHG is a CDK6-dependent lineage, CDK6 is a fate/differentiation gene; inhibition collapses the OPC-/progenitor-like state (Filbin CRISPR, npj Precis Oncol 2025). Escape is predominantly non-genetic: quiescent/persister cells, OPC↔AC↔mesenchymal state-switching, and reversible senescence/SASP (drug-withdrawal regrowth).
Convergent effector = cyclin E1–CDK2 reactivation; plus RTK/PI3K bypass and CDK6-amp drug-titration. RB-loss is the classic route; RB status here is unproven (no formal RB IHC).

Exploit with combinations

+ PI3K/mTOR (everolimus), best CNS-deliverable, G34-inclusive (NCT05843253): blocks the dominant bypass while the CDK6 backbone holds the lineage dependency. strongest available
+ CDK2-selective inhibitor, the true anti-escape move (kills the cyclin E–CDK2 effector) but no CNS-penetrant CDK2i exists, keep generic, not actionable yet. mechanism
+ CDK7 / senolytic (navitoclax), speculative, no glioma/CNS data. speculative

5 · PDGFRA, driver, resistance & the honest combination

Driver vs this patient's variant

PDGFRA amp/activation is a powerful recurrent G34/pHGG lineage co-driver (OPC-like compartment), avapritinib gave 3/7 responses in PDGFRA-altered HGG (Cancer Cell 2025).
But the detected alteration is Y288C, an ER-trapped extracellular mutant, ligand/kinase-conformation-independent, resistant to the PDGFR TKIs tested by Ip 2018 (avapritinib not tested, Ip 2018 recommended evaluating it; futility mechanistically plausible but unproven); antibody/ADC futile (intracellular + BBB). Sensitive to PI3K/mTOR & MEK (Ip 2018).

Escape & the defensible move

Escape: RTK switching (MET/EGFR/FGFR) + OPC↔mesenchymal plasticity, evades single-RTK blockade non-genetically.
Defensible PDGFRA-axis intervention = downstream PI3K-mTOR ± MEK (same spine as item 2). Avapritinib add-on only if a concurrent conformation-competent PDGFRA amplification is demonstrated. Co-RTK (MET/FGFR) pre-emption is mechanism-only. MODERATE

Caveats, read before discussion

Most CDK4/6i and PDGFRA resistance data are breast/ovarian-derived; G34/CNS-specific evidence is limited to one PDX/CRISPR program plus early trials. CDK2i and senolytics have no CNS-penetration or glioma efficacy data. This is provisional, evidence-graded decision-support for MDT discussion, not a protocol. The two named monitoring biomarkers (PDGFRA Y288C, CCNE1) must be added to the SiMSen personalised panel for the 6-weekly cadence to function.

Candidate menu, full drug & combination space

The full set of agents considered, with the anchors identified. CNS: ✓ brain-penetrant · ~ marginal · ✗ poor. Ev = evidence in this patient: supported / hypothesis / speculative / failed/exclude.

Cell cycle, CDK6-amp, strongest truncal target

Drug	CNS	Ev	Note
Abemaciclib	✓~	supported	Current backbone
Ribociclib	✓	supported	CDK6-selective; ~18-mo SD G34 precedent
Palbociclib	✗	exclude	Poor CNS
CDK2i, tagtociclib / BLU-222 / INX-315	✗	hypothesis	True anti-escape move; no CNS agent yet
CDK7i / CDK9i · CDK6 PROTAC	—	speculative	Preclinical only

PI3K / AKT / mTOR, convergent downstream node

Drug	CNS	Ev	Note
Everolimus	✓~	supported	G34-inclusive trial NCT05843253; spine partner
Paxalisib	✓	supported	Brain-penetrant PI3K/mTOR; pHGG programmes
Capivasertib (AKT) · alpelisib (PI3K)	~	hypothesis	Downstream alternatives

MAPK (MEK / RAF / FAK) & PDGFRA / RTK

Drug	CNS	Ev	Note
Avutometinib + defactinib (MEK/RAF + FAK)	✓	supported	NCT06630260, explicitly DHG H3 G34
Trametinib / mirdametinib / selumetinib	~	hypothesis	MEK alternatives
Avapritinib	✓	conditional	Add-on only if conformation-competent PDGFRA amp shown, for Y288C alone, resistance shown for the PDGFR TKIs Ip 2018 tested (avapritinib untested; resolve empirically)
Dasatinib	✗	failed	No responses in pHGG
MET, capmatinib / tepotinib / vebreltinib	~	hypothesis	Pending CSF RNA-seq (Belay MET-amp uninterpretable)

DNA repair / replication stress, truncal ATRX-loss / ALT

Drug	CNS	Ev	Note
Niraparib (PARPi)	✓	supported	Best CNS PARPi; ATRX-mutant glioma trials
Pamiparib / olaparib	✓ / ~	supported	Alternatives (olaparib weaker CNS)
Intrathecal–Ommaya topotecan	✓ (IT)	supported	Only CSF-delivery-validated agent
WEE1i, Debio 0123	✓	supported	Brain-penetrant; adavosertib ✗ excluded
ATRi, elimusertib / ceralasertib	~	hypothesis	ALT synthetic-lethality; camonsertib ✗ no CNS
CHK1i, prexasertib	~	hypothesis	Redundant with WEE1i/ATRi, one only
Carboplatin	~	hypothesis	Exploits ERCC1/NER-deficiency, under-used specific vulnerability
Temozolomide	✓	deprioritise / exclude	Hypermutation risk. MMR-loss + MGMT-methylated + prior TMZ + rising TMB → MMR-loss converts TMZ to a hypermutagen (SBS11; not checkpoint-responsive); MGMT-methylation alone insufficient. Do not re-challenge. Decisive test: MMR IHC + MSI + SBS11 signature on current tissue (see Drug-screen tab).
Pemetrexed ★	~	low	Antifolate; moderate in screen (0.9–3 µM); limited CNS penetration
Paclitaxel ★	✗	delivery-limited	Very potent in reference lines but poor BBB + concordant TUBB3/STMN1 taxane-resistance biology in this tumour, low translatability
Irinotecan ★	~	low	TopoI; moderate (0.9–2.4 µM); SN-38 CNS-limited; historically only in bevacizumab combinations
Daunorubicin ★ / Doxorubicin	✗	delivery-limited	Very potent anthracyclines in screen (deep-blue cytotoxic) but do not cross the BBB, not deliverable to CNS/LMD
Bortezomib ★ → marizomib	✗ → ✓	concept	Bortezomib is the single most potent screen hit but CNS-impenetrant; pursue via brain-penetrant marizomib (prior GBM trials), proteasome dependency worth testing on Calvin's own line
Lomustine/CCNU	✓	failed	5 cycles, progressed; note: nitrosourea crosslinker, does NOT drive the SBS11 hypermutator phenotype (distinct mechanism from TMZ)

XPO1, p53 axis, immune & other

Drug	CNS	Ev	Note
Selinexor (XPO1i)	~	priority experimental, see card	Genuine modest signal: cross-platform-concordant NF-κB/IκBα (NFKBIA) context matches selinexor's core mechanism + reproducible rGBM activity + BBB penetration + paediatric-HGG trial NCT05099003. Not the fusion (passenger) or XPO1-expression (short-read-only).
MDM4i / p53 reactivators (APR-246, ALRN-6924, rezatapopt)	—	not actionable	Mutant p53, see Pre-therapy item 11; reframed as a WEE1/CHK1 node instead
Atorvastatin / simvastatin	~	speculative	GOF-mutant-p53 mevalonate adjunct only
Ad-TD-nsIL12 oncolytic virus	n/a	active	Continue (treating team)
CAR-T / PRAME-TCR-T	—	failed / hold	CAR-T ×4 progressed; G4 lymphopenia, hold all T-cell until recovery
ONC201 / ONC206	✓	discontinue	Empirically failed in this patient; no G34 data

Selinexor (XPO1 inhibitor), assessment

Selinexor blocks XPO1-mediated nuclear export → forced nuclear retention of IκB / FOXO / p53 / p21 and reduced oncoprotein translation. Is there a signal here? Yes, modest but genuine, though not where it was first claimed. It is not the XPO1:PPP4R3B call (single, likely-passenger; does not establish XPO1 dependency) and not XPO1 over-expression (Caris Z +3.71 rk 49 but long-read Z −0.03 / logFC vs GTEx −0.53, a short-read-only signal that fails the item 7 cross-platform gate, like FGFR/PDGFRA). The genuine signals are: (1) a cross-platform-concordant NF-κB / IκBα context, NFKBIA concordantly high (Caris +3.55 rk 65, long-read +2.02, logFC vs GTEx +1.10), with NFE2L2/NRF2 high (logFC +1.90) and RAN concordant-high, and IκBα is a canonical XPO1 cargo, so selinexor's best-validated mechanism (nuclear IκBα retention → NF-κB suppression) maps onto a robust feature of this tumour; (2) a real disease-level signal, reproducible single-agent activity in recurrent HGG (durable RANO responses; KING/NCT01986348), measured BBB penetration (~105 nmol/L in resected tumour), and an active paediatric / AYA HGG trial (selinexor + RT, NCT05099003). Honest tier, exploratory, evidence-graded: rGBM single-agent efficacy is modest overall (PFS6 ~10–17%); the p53-retention arm of the mechanism is blunted by truncal mutant TP53 R273C (p53-independent and NF-κB/FOXO activity retained, so not nullified, unlike MDM4i, item 11); no CSF/leptomeningeal-specific selinexor data (parenchymal exposure only, uncertain for LMD). Verdict, priority experimental arm. The cross-platform-concordant NF-κB / IκBα context gives selinexor a genuine biological anchor in this tumour (its core mechanism, not a generic rationale), reinforced by reproducible recurrent-HGG single-agent activity, measured CNS penetration, and active trial infrastructure (NCT05099003). It should be brought forward, a reasonable Block-2 intensification partner or an early Block-3 switch / trial referral, ranked above the other experimental options, not held back as a last resort. It does not displace the CDK6 / replication-stress spine, but it is now a prioritised candidate within the experimental tier.

Combinations (tiered)

Anchor, start here

Abemaciclib (or ribociclib) + everolimus, CDK6 dependency + CDK4/6i-escape + PDGFRA-Y288C downstream, one CNS-deliverable G34 spine
Spine + avutometinib/defactinib, intensification / 2nd downstream node

On progression, switch, not stack

Selinexor ± RT, priority experimental switch: biologically anchored (cross-platform-concordant NF-κB/IκBα) + reproducible rGBM activity + CNS-penetrant + active trial (NCT05099003)
Niraparib + IT-topotecan (post-CSI, count-gated), ALT/replication-stress backbone
Debio 0123 (WEE1i) single checkpoint node, or elimusertib (ATRi), pick ONE
ATRi + low-dose PARPi, ALT synthetic lethality (sequential)

Experimental / when available

CDK4/6i + CDK2i, true anti-escape (needs a CNS-penetrant CDK2i)
Add-on avapritinib + downstream, only if conformation-competent PDGFRA amp confirmed
Carboplatin, standalone NER-deficiency exploitation

Do not combine / exclude

WEE1i + (niraparib+topotecan), myelotoxicity · WEE1i + ATRi + CHK1i, redundant (one only) · palbociclib / adavosertib / camonsertib, fail CNS gate · T-cell therapy, until lymphocyte recovery · ONC201/206, dasatinib, failed.

Screened & excluded

Genes that appear actionable on Caris short-read but fail the cross-platform / biology gate (item 7). Caris Z = vs Caris internal pan-tumour cohort; LR Z = vs 17 GTEx normal-brain; logFC = long-read vs GTEx.

Gene	Caris Z	LR Z	ΔZ	logFC vs brain	Why excluded
FGFR1	+3.55	−0.62	+4.17	−1.83	Short-read-only; long-read below normal brain; no fusion/mutation
FGFR2	+3.62	−0.12	+3.74	−3.84	Short-read-only; long-read far below brain; no fusion/mutation
NTRK3	+3.29	−0.16	+3.46	−3.45	Short-read-only artifact
TACC3	+2.25	−1.05	+3.31	−0.66	Short-read-only; no FGFR3-TACC3 fusion
NTRK2	+4.26	+2.27	+1.99	+2.23	Survives to long-read, but NTRK is druggable only as a fusion; none present, high TrkB = neural/OPC-lineage expression, not a target

Decider

No NTRK or FGFR fusion or activating mutation anywhere in the case (only real fusions: CDK6:VIPR2, XPO1:PPP4R3B). Without a fusion/hotspot, neither family is actionable regardless of expression. Larotrectinib/entrectinib (NTRK) and FGFR inhibitors are not indicated. This entry validates the item 7 concordance gate, it correctly screens these out.

Why Caris short-read & UCLA long-read disagree for FGFR, methodology

Different denominators (the biggest reason)

Caris MI Tumor Seek is hybrid-capture (exome-anchored) FFPE short-read RNA, and its "high" call is a rank/Z within Caris's own pan-tumour FFPE cohort, not vs normal brain. "High Caris percentile for FGFR1/FGFR2" means high relative to other tumours on that platform, a different question from over-expressed vs tissue-of-origin. The long-read Z is vs GTEx normal brain. Different reference → apparent discordance with no biological over-expression.

Gene-family multi-mapping & capture bias

FGFR1–4 and NTRK1–3 are highly homologous (plus pseudogenes); short reads + capture baits cross-hybridise and redistribute reads across family members, inflating locus counts. The signature here is diagnostic: NTRK3/TACC3 discordant while the distinct NTRK2 is concordant, family-level cross-mapping, not biology. FFPE degradation and FGFR2 IIIb/IIIc alternative splicing compound it.

Unit mismatch & long-read undercounting

Caris TPM (length-normalised) vs long-read CPM (not) systematically biases long genes like FGFR across platforms. And long-read (Nanopore/Iso-Seq) has far lower depth, it under-calls low/mid-expression and degraded/long transcripts (dropout, 5′-truncation). So ~1500 TPM vs ~2 CPM overstates the true gap; long-read is not ground truth either.

Cross-platform caveat & lesson

GTEx is itself short-read Illumina, so the long-read-CPM-vs-GTEx logFC (−1.8 to −3.8) is also cross-platform, not a clean tumour-vs-normal contrast. Net: truth for FGFR1/2 is most likely "not strongly over-expressed vs brain", but neither platform fixes the absolute level, needs IHC/RNA-ISH if it ever mattered (it doesn't, absent a fusion). General rule: a Caris internal-cohort percentile is a relative pan-tumour rank on a capture FFPE short-read platform, for homologous families (FGFR/NTRK) in n=1, give it low standalone weight, hypothesis-generating only. Refs: Iwabuchi 2026 (10.1093/narmme/ugag006), Cabanski 2014 (10.1016/j.jmoldx.2014.03.004), Priedigkeit 2021 (10.1186/s13058-020-01379-3); Caris validation Oncotarget art. 28761 (vendor-disclosed).

What weight should the Caris internal cohort carry?, the random-vs-systematic distinction

The Caris reference cohort is large (tens of thousands) and the long-read-vs-GTEx contrast is itself cross-platform, which raises the question of how heavily the Caris percentile should weigh. The key principle: cohort size reduces RANDOM error, not SYSTEMATIC error. With n in the tens of thousands the percentile rank itself is statistically robust, a genuine strength. However, FGFR/NTRK paralog cross-mapping + capture bias is a systematic distortion intrinsic to the assay: the reference and the patient sample are shifted by the same offset, so a larger cohort yields greater precision around a value that is systematically biased for this gene family, n reduces noise, not bias. Equally, the long-read/GTEx contrast is itself imperfect (n=17, cross-platform, undercounts), it flags uncertainty rather than proving FGFR is low. Practical weighting: give the Caris percentile HIGH weight when (a) the reference is lineage/glioma-specific, (b) the gene is uniquely mappable (not a homologous family), (c) corroborated by protein/DNA, (d) the drug's evidence base is expression-selected; give it LOW standalone weight when any fail, for FGFR/NTRK (b)+(c)+(d) all fail at once. The embedded normalisation here is a single global mean/SD (pan-tumour). The glioma-lineage-specific Caris percentile and an orthogonal FGFR assay (IHC/RNA-ISH/fusion) are the resolving tests; for FGFR, none of it changes the call without a fusion (absent here).

Cell line screening

Baseline single-agent screen (Misvik) across reference glioma cell models. The lineage baseline; for the patient-specific reads see the Filbin GBX63 PDX tab.

Cell-model panel

Reference models, screened (this dataset)

KNS42, paediatric H3F3A G34V GBM line; the closest available lineage model to this tumour
MISB547 · MISB577 · MISB592, Misvik glioma models
84 agents, 5-point single-concentration response + fitted IC50

Filbin GBX63 PDX (Calvin's patient line)

Patient-derived PDX line ex-vivo drug response (Nov 2025) referenced throughout the axis tabs · full data, source figures and analysis in the Filbin GBX63 PDX tab

Patient-derived line (Filbin GBX63)

Calvin's PDX-derived line was established and screened by the Filbin Lab (report dated 03 Nov 2025). Full data and analysis are in the Filbin GBX63 PDX tab.
The Misvik panel below remains the lineage-reference baseline; the patient line is the orthogonal patient-specific read.

How to read the screen

Heatmap

For each cell line the drug row shows several cells = individual test concentrations, with the final column per line = the mean across those concentrations. Colour encodes the kind of effect: deep blue = growth inhibition / cell kill (toward −100 ≈ up to 100% kill, cytotoxic); white = no effect in either direction; red = net growth (drug not inhibiting). Read the row colour first, it shows what is happening biologically.

IC50 sheet

The IC50 sheet gives one number per drug per cell line: the concentration, in µM, at which the fitted dose-response curve crosses 50% of its own maximum observed response. It therefore encodes potency, not the biological nature of the effect. The heatmap should be read first, because it tells us what is happening biologically: cytotoxicity, growth arrest/stalling, or little/no effect. The IC50 should then be read second, because it tells us the concentration at which that effect occurs. Units are µM throughout, e.g. 0.036 µM = 36 nM (highly potent), 22 µM is modest/weak, 344 µM is essentially inactive in practical terms. "n/r" / "+infinity" means the fitted curve did not reach 50% of its maximum response within the tested concentration range; this does not by itself distinguish "inactive at clinically achievable doses" from "inactive within the tested dose range". A low/potent-looking IC50 must be interpreted against the heatmap: a low IC50 with a pale or minimally responsive row may simply reflect curve-fitting noise or extrapolation rather than meaningful activity (e.g. telmisartan), whereas a low IC50 with a deep-blue row is much more likely to represent genuine cytotoxicity (e.g. vincristine ≈ 250 nM). The key unresolved caveat is that the absolute concentrations behind the heatmap columns have not yet been provided; these IC50s therefore cannot yet be anchored to clinically relevant exposure, especially free-brain Cmax, and "n/r" should not yet be interpreted as "inactive at clinically achievable exposure". For now the results should be treated as relative potency plus effect-type within this assay, pending the concentration key and clinical-exposure mapping.

Baseline screen heatmap

Scale, figure legend: % growth inhibition.deep blue (−100) = maximal growth-inhibition / cell-kill· white (0) = no effect· red (+100) = no inhibition / net growth— empirical orientation cross-checked against IC50 (most-potent Bortezomib/Vincristine are the deep-blue rows; inactive Miglustat/Telmisartan are pale/red); confirm convention with Misvik.

Top hits, potent & cytotoxic across the panel

Most potent (min IC50 < 2 µM, ≥3 lines), heatmap-cytotoxic

Bortezomib (20S proteasome (β5 subunit, reversible)), min IC50 36 nM
Panobinostat (HDAC I · II · IV), min IC50 77 nM
Daunorubicin (Topoisomerase II), min IC50 107 nM
Dasatinib (BCR-ABL · SRC family · PDGFR · KIT · EPHA2), min IC50 114 nM
Tegatrabetan (BC2059) (β-catenin (CTNNB1) degrader), min IC50 122 nM
Doxorubicin (Topoisomerase I/II), min IC50 144 nM
Vincristine (Microtubule), min IC50 179 nM
Paclitaxel (None), min IC50 231 nM
Selinexor (XPO1 (CRM1)), min IC50 374 nM
Topotecan (Topoisomerase I), min IC50 381 nM
Tepotinib (cMET, PDE3 · PKA), min IC50 426 nM
Methotrexate (DHFR), min IC50 432 nM
Lorlatinib (ALK · ROS1), min IC50 443 nM
IAG933 (TEAD (pan-isoform, PPI inhibitor)), min IC50 518 nM
Cabozantinib (MET · VEGFR2 · AXL · RET · KIT), min IC50 654 nM
Pevonedistat (MLN4924) (NEDD8-activating enzyme (NAE, UBA3)), min IC50 763 nM
Repotrectinib (TRKA/B/C · ROS1 · ALK), min IC50 876 nM
Irinotecan (Topoisomerase I), min IC50 895 nM

Case interpretation, read with caveats

Selinexor (XPO1), sub-µM–~1.7 µM, deep-blue cytotoxic on every line. Independently consistent with the cross-platform-concordant NF-κB/IκBα signal (Candidate menu / Genomics); supports its elevation to a priority experimental arm.
Bortezomib, the single most potent hit (36–125 nM, deep blue all lines) but CNS-impenetrant; the brain-penetrant proteasome inhibitor marizomib is the translatable analogue (prior GBM trials).
Avapritinib, weak in these G34 reference lines (14–36 µM, pale). Tempers (does not support) avapritinib on lineage grounds, though these models are not Y288C-bearing and not the patient's cells; the on-treatment SiMSen burden remains the arbiter.
Temozolomide, weak in vitro (62–126 µM); short-assay IC50 under-represents TMZ, but offers no positive signal here. See hypermutation caveat below.
Abemaciclib > ribociclib (3–12 vs 21–115 µM), consistent with abemaciclib as the CDK4/6 choice.
TUBB3 caveat: paclitaxel/vincristine/daunorubicin are deeply cytotoxic in these lines, but Calvin's tumour shows cross-platform-concordant TUBB3 (class III β-tubulin, Caris +1.65 / long-read +2.19 / logFC vs GTEx +1.45) and high STMN1, established taxane/vinca-resistance biology, so these hits may not transfer to his tumour; his own line will test this directly.

Drugs raised for discussion, with intrathecal-delivery assessment

For leptomeningeal disease the relevant question is not only systemic BBB penetration but whether the agent can be given intrathecally / intra-Ommaya.

★ Pemetrexed (TYMS/DHFR), screen: moderate (0.9–3 µM, n/r MISB577); systemic CNS limited. IT: a credible, increasingly evidence-based option, phase 1/2 in NSCLC-LM established 50 mg IT + dexamethasone with ~80% RANO response and median OS ~12 mo (Fan 2021/2024); dominant toxicity reversible myelosuppression, arachnoiditis mitigated by dexamethasone. The strongest IT candidate among these, discuss as a CSF-directed adjunct.
★ Paclitaxel, very potent cytotoxic in reference lines (231–393 nM) but poor BBB and concordant TUBB3/STMN1 taxane-resistance biology in this tumour. IT: only older/anecdotal phase-I data for neoplastic meningitis (Cremophor vehicle → arachnoiditis); never reached standard practice. Low translatability.
★ Irinotecan (TopoI), moderate (0.9–2.4 µM); SN-38 CNS-limited. IT: no credible human data (only systemic case reports), not a current option.
★ Daunorubicin / doxorubicin (anthracyclines), very potent (107–280 nM, deep blue) but do not cross the BBB and are neurotoxic intrathecally, effectively contraindicated IT. Not deliverable to CNS/LMD by any route.
★ Bortezomib, the single most potent screen hit but CNS-impenetrant and no IT data with a neurotoxicity signal; pursue the proteasome concept via brain-penetrant marizomib (prior GBM trials), not bortezomib, and test on Calvin’s own line.

Net: the most potent screen classes (proteasome, anthracycline, taxane) are largely delivery-limited systemically and not IT-deliverable. The deliverable exceptions for LMD are intrathecal topotecan (established), intrathecal pemetrexed (credible emerging adjunct), the established IT comparators (MTX, cytarabine/liposomal-cytarabine, thiotepa, etoposide), and the systemically-anchored selinexor and CDK4/6 / replication-stress agents.

Full IC50 table, by drug class

Values are IC50 in µM, shown as potency chips: very potent <0.5 potent 0.5–2 moderate 2–10 weak 10–50 n/r · inactive. ★ = raised for discussion. The potency band reflects the lowest IC50 across the four lines.

Drug	Target	KNS42	MISB547	MISB577	MISB592	potency
Proteostasis / oxidative
Bortezomib ★	20S proteasome (β5 subunit, reversible)	113 nM	36 nM	105 nM	125 nM	very potent
Selinexor	XPO1 (CRM1)	1.69 µM	374 nM	1.22 µM	1.29 µM	very potent
Pevonedistat (MLN4924)	NEDD8-activating enzyme (NAE, UBA3)	763 nM	1.38 µM	1.78 µM	18 µM	potent
Auranofin	TXNRD1	1.03 µM	1.11 µM	1.03 µM	1.37 µM	potent
Pimitespib (TAS-116)	HSP90α / β	15 µM	29 µM	11 µM	22 µM	weak
Epigenetic / transcription
Panobinostat	HDAC I · II · IV	189 nM	77 nM	159 nM	217 nM	very potent
Tegatrabetan (BC2059)	β-catenin (CTNNB1) degrader	362 nM	125 nM	122 nM	294 nM	very potent
IAG933	TEAD (pan-isoform, PPI inhibitor)	5.38 µM	1.58 µM	701 nM	518 nM	potent
Fimepinostat (CUDC-907)	HDAC I/II + PI3Kα	2.21 µM	979 nM	1.51 µM	1.75 µM	potent
Pelabresib (CPI-0610)	BRD2/3/4	17 µM	9.73 µM	5.83 µM	9.96 µM	moderate
GSK3326595 (Pemrametostat)	PRMT5	17 µM	10 µM	37 µM	22 µM	weak
Revumenib	Menin – KMT2A interaction	94 µM	n/r	60 µM	16 µM	weak
Valemetostat	EZH1 · EZH2 (dual)	n/r	n/r	40 µM	n/r	weak
Cytotoxic chemotherapy
Daunorubicin ★	Topoisomerase II	280 nM	206 nM	171 nM	107 nM	very potent
Doxorubicin	Topoisomerase I/II	504 nM	438 nM	457 nM	144 nM	very potent
Vincristine	Microtubule	322 nM	245 nM	179 nM	249 nM	very potent
Paclitaxel ★	—	269 nM	393 nM	231 nM	279 nM	very potent
Topotecan	Topoisomerase I	831 nM	470 nM	396 nM	381 nM	very potent
Methotrexate	DHFR	502 nM	894 nM	1.08 µM	432 nM	very potent
Irinotecan ★	Topoisomerase I	2.35 µM	1.11 µM	1.77 µM	895 nM	potent
Pemetrexed ★	TYMS · DHFR · GART	1.57 µM	908 nM	n/r	2.99 µM	potent
Cisplatin	DNA crosslinks	9.90 µM	4.34 µM	18 µM	13 µM	moderate
5-FU (5-Fluorouracil)	TYMS (FdUMP)	88 µM	35 µM	73 µM	139 µM	weak
Carboplatin	Inter/intrastrand DNA crosslinks	189 µM	75 µM	132 µM	61 µM	inactive / n-r
Temozolomide	O6-MeG	126 µM	92 µM	106 µM	62 µM	inactive / n-r
Lomustine (CCNU)	DNA alkylation / carbamoylation / interstrand crosslinks	142 µM	79 µM	89 µM	88 µM	inactive / n-r
Carmustine (BCNU)	DNA alkylation / carbamoylation	344 µM	211 µM	105 µM	98 µM	inactive / n-r
Cell-cycle / DDR
Adavosertib (AZD1775)	WEE1	2.41 µM	1.05 µM	8.61 µM	1.59 µM	potent
Alisertib (MLN8237)	AURKA (Aurora A)	4.49 µM	3.93 µM	4.20 µM	2.28 µM	moderate
Abemaciclib	CDK4/6	9.89 µM	12 µM	3.73 µM	2.30 µM	moderate
Ceralasertib (AZD6738)	ATR	6.54 µM	2.97 µM	4.40 µM	26 µM	moderate
Pamiparib	PARP	12 µM	8.14 µM	4.65 µM	6.55 µM	moderate
Niraparib	PARP1/2	15 µM	20 µM	36 µM	35 µM	weak
Tagtociclib (PF-07104091)	CDK2 · Cyclin E	29 µM	18 µM	41 µM	55 µM	weak
Ribociclib	CDK4/6	36 µM	115 µM	n/r	21 µM	weak
Novobiocin	POLQ (ATPase)	n/r	n/r	127 µM	63 µM	inactive / n-r
RTK / kinase
Dasatinib	BCR-ABL · SRC family · PDGFR · KIT · EPHA2	1.87 µM	313 nM	190 nM	114 nM	very potent
Tepotinib	cMET, PDE3 · PKA	1.35 µM	1.40 µM	1.28 µM	426 nM	very potent
Lorlatinib	ALK · ROS1	2.50 µM	2.08 µM	1.01 µM	443 nM	very potent
Cabozantinib	MET · VEGFR2 · AXL · RET · KIT	4.31 µM	3.88 µM	654 nM	660 nM	potent
Repotrectinib	TRKA/B/C · ROS1 · ALK	4.16 µM	2.56 µM	876 nM	1.07 µM	potent
Gilteritinib	AXL · FLT3 (ITD and TKD mutants)	9.25 µM	30 µM	1.88 µM	1.08 µM	potent
Furmonertinib	EGFR (pan-mutant)	9.55 µM	6.72 µM	1.83 µM	1.12 µM	potent
Entrectinib	TRKA/B/C · ROS1 · ALK	7.36 µM	4.53 µM	1.33 µM	2.25 µM	potent
Osimertinib	EGFR (T790M-preferential, mutant)	15 µM	63 µM	8.73 µM	1.96 µM	potent
Defactinib	FAK (PTK2)	6.89 µM	4.38 µM	2.72 µM	2.12 µM	moderate
Pyrotinib	EGFR · HER2 · HER4	4.78 µM	11 µM	3.37 µM	3.35 µM	moderate
Neratinib	EGFR · HER2 · HER4 (irreversible)	18 µM	17 µM	7.37 µM	4.25 µM	moderate
Dacomitinib	EGFR · HER2 · HER4 (irreversible)	7.46 µM	7.91 µM	6.68 µM	5.15 µM	moderate
Ponatinib	BCR-ABL · FGFR1-4 · PDGFRA/B · VEGFR · SRC · FLT3	15 µM	6.92 µM	5.21 µM	5.63 µM	moderate
Futibatinib	FGFR1 – 4 (covalent, irreversible)	n/r	37 µM	5.59 µM	5.49 µM	moderate
Pemigatinib	FGFR1 – 3	n/r	n/r	12 µM	53 µM	weak
Avapritinib	PDGFRA (D842V / exon 18) · KIT	22 µM	36 µM	18 µM	14 µM	weak
Lenvatinib	FGFR1-4 · VEGFR1-3 · PDGFRα · RET · KIT	30 µM	27 µM	n/r	n/r	weak
Pexidartinib	CSF1R · KIT · FLT3	n/r	n/r	63 µM	64 µM	inactive / n-r
PI3K/AKT/mTOR/MAPK
Mirdametinib	MEK1/2	5.72 µM	4.78 µM	10 µM	908 nM	potent
Everolimus	mTORC1	1.86 µM	1.48 µM	1.61 µM	8.15 µM	potent
Trametinib	MEK1/2	2.63 µM	1.61 µM	7.38 µM	1.61 µM	potent
Avutometinib (VS-6766)	RAF, MEK complex (clamp)	8.52 µM	45 µM	3.37 µM	5.46 µM	moderate
Capivasertib	AKT1/2/3	12 µM	3.85 µM	n/r	6.91 µM	moderate
Paxalisib	PI3K (p110α/δ) · mTORC1/2	37 µM	20 µM	9.66 µM	7.08 µM	moderate
Ruxolitinib	JAK1 · JAK2	10 µM	27 µM	n/r	90 µM	weak
Azelnidipine	SIRPα · PVR · MEK1/2 (dual innate-checkpoint + MAPK)	14 µM	14 µM	48 µM	13 µM	weak
p53 / lineage / other targeted
APR-246 (Eprenetapopt)	Mutant p53 (R273, Y220, R175 classes)	13 µM	7.12 µM	3.80 µM	3.25 µM	moderate
Retinoic acid	RARα · RARβ · RARγ	8.48 µM	3.32 µM	4.98 µM	6.68 µM	moderate
ONC201 (Dordaviprone)	DRD2 · ClpP	36 µM	75 µM	14 µM	8.13 µM	moderate
Nirogacestat	γ-secretase	n/r	n/r	n/r	24 µM	weak
Tetrac	Integrin αvβ3 at T4 binding site	n/r	n/r	n/r	28 µM	weak
Repurposed / metabolic / other
Pitavastatin	HMG-CoA reductase	2.05 µM	7.59 µM	5.06 µM	30 µM	moderate
Perhexiline	CPT1 inhibitor · FAO	10 µM	10 µM	3.01 µM	36 µM	moderate
Riluzole	Glutamate release · mGluR3 · Na channels	11 µM	14 µM	7.71 µM	3.99 µM	moderate
Vortioxetine	SERT · 5-HT · lysosomal destab	93 µM	5.86 µM	7.18 µM	7.26 µM	moderate
Tamoxifen	SERM · PKC inhibitor	39 µM	46 µM	28 µM	14 µM	weak
Ibudilast	PDE4 · MIF antagonism	125 µM	59 µM	14 µM	14 µM	weak
Desloratadine	H1 antagonist · cationic amphiphilic drug	46 µM	31 µM	28 µM	15 µM	weak
Chlorpromazine	DRD2 · autophagy	45 µM	17 µM	20 µM	16 µM	weak
Aprepitant	TACR1 (NK-1 receptor)	89 µM	68 µM	n/r	18 µM	weak
Telaglenastat (CB-839)		n/r	45 µM	51 µM	26 µM	weak
Propranolol	β1/β2 antagonist	n/r	n/r	26 µM	58 µM	weak
DFMO (Eflornithine)	ODC1	147 µM	n/r	37 µM	n/r	weak
Ranolazine	Late-Na · partial pFAO	43 µM	100 µM	78 µM	38 µM	weak
Nitazoxanide	OXPHOS · Wnt · autophagy	70 µM	46 µM	59 µM	40 µM	weak
Bazedoxifene	SERM · IL-6/GP130 inhibitor	46 µM	52 µM	n/r	65 µM	weak
Miglustat	Glucosylceramide synthase	n/r	n/r	51 µM	n/r	inactive / n-r
Telmisartan	AT1R · PPARγ partial agonist	105 µM	72 µM	n/r	n/r	inactive / n-r
Ketorolac	COX-1/2 · Rac1/Cdc42	n/r	n/r	n/r	n/r	inactive / n-r

Temozolomide, hypermutation caveat (decisive for menu placement)

The glioma hypermutation literature is explicit here. With MGMT-methylated + MMR-protein loss (MSH2/MSH6 ND) + prior TMZ ×6 + re-challenge + rising TMB + progression + LMD spread, this is the high-risk archetype for TMZ-induced hypermutation: loss of MMR converts TMZ from a cytotoxic into a hypermutagenic, resistance-promoting agent (SBS11; MSI-stable on bulk testing), and treatment-induced hypermutation in glioma does not become checkpoint-responsive (Touat 2020 Nature PMID 32322066; Johnson 2014 Science PMID 24336570; van Thuijl 2015 PMID 25724300; Yu 2021 PMID 33823014). MGMT-methylation alone does not justify TMZ once MMR is lost. Recommendation: remove TMZ from active consideration / place at the very bottom and do not re-challenge; decisive test = MMR IHC (MLH1/PMS2/MSH2/MSH6) + MSI + SBS11 mutational-signature on current tissue. Note CCNU/BCNU act via interstrand crosslinks (distinct mechanism) and do not drive the SBS11 hypermutator phenotype, a different risk/resistance profile, not "safe".

Filbin GBX63 PDX · Dana-Farber / Boston Children's

Ex-vivo drug screen performed by the Filbin Lab (Dana-Farber / Boston Children's Brain Tumor Center) on GBX63, Calvin's patient-derived cell line established from his resection. Formal report dated 03 Nov 2025.

Filbin GBX63 PDX · Dana-Farber / Boston Children's · Nov 2025

GBX63 is Calvin's PDX-derived cell line established from his resection by the Filbin Lab. The Nov 2025 package below covers a 49-compound single-agent screen and three Ribociclib-anchored combination matrices on this line.

Methodology

09 Sep 2025: PDX cells thawed, plated in 5 mL ULA flask + laminin
21 Oct 2025: Single-agent screen, 49 compounds, 2,000 cells/well in 384-well non-treated plate; drugged twice over 7 days; CellTiter-Glo viability
21 Oct 2025: Combination screen, three Ribociclib-anchored pairs; 4×4 dose grid 10 / 100 / 1000 / 2000 nM each axis
Loewe synergy score analysis · 3D dose-surface + dose-pair matrix

Reading the screen

RelVia = relative viability (1.0 = untreated control, 0 = full kill)
Single-agent: two-dose single-concentration (2 µM and 10 µM); no IC50 from this dataset
Combinations: 4 × 4 dose grid (10 / 100 / 1000 / 2000 nM each axis) with Loewe synergy scoring; read the dose-pair matrix and the 3D global mean / p-value together
The PDX line carries the truncal genotype (H3.3 G34R · TP53 R273C · ATRX-loss · CDK6 amp). Subclonal CSF-emergent variants (e.g. PDGFRA Y288C) may or may not have been captured during PDX establishment; targeted re-genotyping of the line is needed to interpret subclonal-variant-dependent hits

Source

Pediatric Neuro-Oncology Precision Medicine Package, Filbin Lab, Longwood Center LC6101, 360 Longwood Avenue, Boston MA 02215. Heatmaps below are redrawn from the source data tables.

Single-agent relative viability · 18 reported compounds (GBX63)

Relative viability at 2 µM and 10 µM. Lower = stronger kill. Strong-kill threshold ≤ 0.20 at 10 µM; near-complete kill at ≤ 0.05.

Single-agent relative viability · 2 µM / 10 µM

Drug	Class	2 µM	10 µM	Read · cross-reference
Selinexor	SINE / XPO1 inhibitor	0.0000	0.0000	Complete kill at both doses. Functional anchor for the NF-κB / XPO1 axis (ax11). Converges with NFKBIA Z +3.55 / RAN Z +2.42.
Afatinib	EGFR (irreversible)	0.0000	0.0000	Complete kill at both doses. EGFR-class active despite ND on MRM-MS. ax4 multi-RTK.
Dacomitinib	EGFR (irreversible)	0.0000	0.0000	Complete kill. Second-generation EGFR confirms the class.
Avapritinib	KIT / PDGFRA TKI	0.0011	0.0000	Complete kill on the patient PDX. ax4 multi-RTK. Calvin's PDGFRA Y288C is subclonal (CSF Feb 2026 28% VAF, declining); whether this PDX line captured the Y288C subclone is not yet known. Genotype the line before reading this as an on-target Y288C result (Ip 2018 documents Y288C resistance to other PDGFR TKIs, avapritinib untested).
Trametinib	MEK1/2	0.0241	0.0000	Complete kill at 10 µM. ax5 vertical cascade.
Pemigatinib	FGFR1–4	0.0212	0.0029	Strong FGFR kill. ax4 multi-RTK. FGFR proteins ND on mProbe; functional data dominates.
Erdafitinib	FGFR1–4	0.0206	0.0046	Strong FGFR kill. ax4.
Futibatinib	FGFR1–4 covalent	0.0632	0.0112	FGFR class triplet all converge on strong kill. Covalent FGFR-trap holds at higher dose.
Entrectinib	NTRK / ROS1 / ALK	0.0350	0.0188	Strong kill despite no NTRK fusion. ax4 (NTRK2 Z +4.26) + RPPA ROS1 46% / TRK activation-loop 39%.
ONC201	DRD2 / ClpP agonist	0.0196	0.0414	Strong kill. Consistent with the current active therapy line.
Niraparib	PARP-i (CNS-active)	0.0611	0.1353	Strong kill at 2 µM, moderate at 10 µM. ax2 DNA-repair effector deficit; functional anchor for the PARP-trap rationale.
Lomustine	nitrosourea alkylator	0.0740	0.0836	Strong kill, dose-flat. Chemotherapy tab.
Pamiparib	PARP-i	0.1851	0.1206	Moderate kill. Reported as "Parimparib" in source.
Larotrectinib	NTRK-fusion-selective	0.1822	0.2480	Moderate kill; entrectinib outperforms (broader spectrum).
Ribociclib	CDK4/6	0.1425	0.2228	Moderate kill as monotherapy. ax1; combination data below.
Palbociclib	CDK4/6	0.4935	0.0649	Strong dose-response. ax1.
Abemaciclib	CDK4/6 + CDK9	0.5864	0.1691	Dose-dependent kill. CDK9 polypharmacology distinguishes it. ax1.
Temozolomide	alkylator chemo	0.3005	0.3783	Weak kill, dose-flat. Read in context of the Chemotherapy tab hypermutation caveat.

Combination screen · three Ribociclib-anchored pairs

4×4 dose matrices and Loewe synergy surfaces. Filbin's threshold: a single dose-pair cell > 5 = local synergy at that pair. The 3D mean and p-value indicate the global surface picture; the two must be read together.

Global synergy summary (3D Loewe surface)

Combination	Global mean	p-value	Peak dose-pair	Read
Ribociclib + Avapritinib	+41.61	0.27	5.68 at Rib 100 nM × Avap 2000 nM	Localised synergy peak; global trend not statistically significant.
Ribociclib + Bumetanide	−161.65	0.002	5.67 at Rib 100 nM × Bume 2000 nM	Globally antagonistic (negative mean, p = 0.002), one local synergy cell. Read with caution.
Ribociclib + Ponatinib	−34.21	0.47	5.91 at Rib 100 nM × Pona 2000 nM	Localised synergy peak; global trend not statistically significant. Cleanest peak of the three.

Ribociclib + Avapritinib · 4×4 dose matrix

Ribociclib + Bumetanide · 4×4 dose matrix

Ribociclib + Ponatinib · 4×4 dose matrix

Single-agent baselines on the combination plate · 4 drugs × 4 doses

Avapritinib, Bumetanide, Ponatinib and Ribociclib at the four combination-plate dose levels (10 / 100 / 1000 / 2000 nM). Provided alongside the combination matrices to anchor the synergy calculations.

Analysis · what the Filbin screen says and where the limits are

Single-agent signals that cross-reference the molecular axes

FGFR class (erdafitinib · futibatinib · pemigatinib) all 0.00x at 10 µM despite FGFR proteins ND on MRM-MS · ax4 multi-RTK; functional dominance over the protein-detection caveat.
EGFR irreversibles (afatinib · dacomitinib) 0.00 / 0.00 · ax4; contradicts the "EGFR IHC+ but not amplified" reservation about target engagement.
Avapritinib 0.00 / 0.00 on the patient PDX · ax4; clean PDGFRA-class kill on Calvin's own cells. Interpretation against the subclonal Y288C variant depends on whether the PDX line carries it (genotyping pending).
Entrectinib 0.035 / 0.019 · ax4; supports NTRK2 Z +4.26 + RPPA ROS1 46% + TRK activation-loop 39%.
Selinexor 0.00 / 0.00 · ax11 NF-κB / XPO1; converges with NFKBIA Z +3.55 / RAN Z +2.42.
Niraparib 0.06 at 2 µM · ax2 DNA-repair effector deficit; functional anchor for the PARP-trap trunk.
Trametinib 0.024 / 0.000 · ax5 cascade (downstream of RAS / RAF).
ONC201 0.02 / 0.04 · current active therapy line.
Lomustine 0.07 / 0.08 · nitrosourea baseline for chemo planning.

Caveats that bound the inference

Two-dose single-concentration screen at 2 µM and 10 µM; no IC50 fits from this dataset, no sub-2 µM resolution.
PDX-establishment subclone capture is not fully characterised. The truncal lineage (H3.3 G34R · TP53 R273C · ATRX-loss · CDK6 amp) will be present, but subclonal CSF-emergent variants (PDGFRA Y288C in particular) may have been lost or enriched during PDX growth. Re-genotype the line to anchor subclonal-variant-dependent reads (especially the avapritinib hit).
Two of the three combinations are not globally synergistic across the dose surface (Bumetanide: mean −162, p = 0.002, globally antagonistic; Ponatinib: mean −34, p = 0.47, not statistically significant). Filbin's >5 cell threshold reports local synergy; the headline values 5.68 / 5.67 / 5.91 are localised peaks at one dose pair, not surface-wide synergy.
The Page-6 combination-plate single-agent baselines read > 1.0 at low doses for Ribociclib, indicating plate-level normalisation differs from the dedicated single-agent screen. The single-agent table is the primary reference; the Page-6 figure is included as the synergy-score anchor.