Eight readouts, one biological network

The cleanest causal arrow in cancer biology is also the most stubbornly resistant to direct exploitation. KRAS-GTP activates the RAF kinases, which phosphorylate MEK1/2, which phosphorylate ERK1/2, which drive proliferation, survival, and lineage programmes. Each node has been drugged. None of the downstream nodes, on their own, have produced a registrational signal in KRAS-mutant pancreatic ductal adenocarcinoma. The Infante Phase 2 of trametinib plus gemcitabine in 1L PDAC reported HR 0.98 — a near-perfect null at hazard ratio 1.0 — with median OS 8.4 versus 6.7 months (p=0.45) ¹. The Chung S1115 trial pairing selumetinib with the AKT inhibitor MK-2206 in 2L PDAC was worse than that: HR 1.37, 3.9 versus 6.7 months, vertical combination underperforming chemotherapy.

These doublets fail for a reason that is not subtle. Acute MEK inhibition releases negative feedback on RTKs; the pathway reactivates within days through RTK-driven RAS-GTP loading (Xue Nature 2020)². Vertical inhibition without simultaneous KRAS engagement closes the loop the same way an engineer would predict. Direct active-state inhibitors change the upstream constraint but not the downstream feedback architecture. MEK combinations remain in development; the historical track record argues for skepticism of pathway-deepening combinations as a clinical class.

PMID anchors: 24915778 · 31915379

KRAS rewires metabolism. Ying et al. Cell 2012 ³ established that oncogenic KRAS in PDAC drives glucose uptake, channels glycolytic intermediates into the non-oxidative pentose phosphate pathway, and decouples ribose biogenesis from NADPH redox control. Son et al. Nature 2013 ⁴ extended the picture: PDAC cells route glutamine through a non-canonical aspartate-oxaloacetate-malate-pyruvate pathway that increases NADPH/NADP+ ratio, and the entire reprogramming is KRAS-dependent. Commisso et al. Nature 2013 ⁵ added a third pillar — Ras-transformed cells acquire extracellular protein through macropinocytosis, internalise it, and proteolytically degrade it into amino acids that enter central carbon metabolism, with pharmacological inhibition of macropinocytosis compromising pancreatic xenograft growth.

The clinical translation has run negative across mTOR, PI3K, and AKT in unselected PDAC; no single downstream node captures the redundancy. The selumetinib + MK-2206 doublet (Chung S1115) tested vertical MEK + AKT suppression in 2L PDAC and read HR 1.37 against chemotherapy. KRAS dependency in PDAC is metabolic at its core, but PI3K, AKT, mTOR, and glutaminase taken one at a time leave the network intact — the redundancy that direct KRAS engagement, in principle, would collapse.

The mTORC1-selective axis is named in earlier RVMD↗ pipeline disclosures as RMC-5552; it does not appear in the Revolution Medicines↗ Q1 2026 10-Q's RAS(ON) pipeline self-summary, leaving its programmatic status in 2026 ambiguous in the public record.

PMID anchors: 22541435 · 23535601 · 23665962 · 25323927

The tri-complex inhibitor mechanism rests on cyclophilin A (CYPA, gene PPIA). Schulze et al. Science 2023 ⁶ established that a small-molecule glue can recruit CYPA to create a neomorphic interface against active-state KRAS-G12C; Holderfield et al. Nature 2024 ⁷ extended the framework to concurrent inhibition of oncogenic and wild-type RAS-GTP through the same chaperone-recruitment principle. CYPA sequestration is therefore the mechanism's foundation — which leaves an unasked question: what does the rest of CYPA's biology do once CYPA is sequestered?

Liu et al. Cell 1991 ⁸ established the canonical answer for the cyclosporin / cyclophilin axis: the cyclophilin–cyclosporin A complex binds and inhibits calcineurin, the Ca²⁺/calmodulin-dependent phosphatase that activates the transcription factor NFAT and drives T-cell-receptor signaling. The two-step coupling — drug binds cyclophilin, complex inhibits calcineurin, calcineurin inhibition silences NFAT — is the mechanistic core of clinical immunosuppression. Schiene-Fischer et al. Angewandte Chemie 2022 ⁹ reviewed the chemical space of non-immunosuppressive cyclophilin inhibitors derived from cyclosporin and sanglifehrin scaffolds; the literature documents that selectivity is achievable — but selectivity must be designed in, and proof of selectivity in a given chemotype is a study, not an assumption.

The peer-reviewed RAS tri-complex literature has not measured whether the chaperone-recruitment chemotypes that engage CYPA on active-state RAS also impair calcineurin / NFAT in clinical pharmacology — neither in the preclinical immune-cell work, nor in the Phase 1 safety reporting. The absence is the finding. Checkpoint-inhibitor combinations depend on T-cell function. CYPA sequestration's downstream effect on calcineurin / NFAT is a measurable axis, and no peer-reviewed clinical study reports the measurement.

PMID anchors: 37590355 · 38589574 · 1715244 · 35290695

KRAS-mutant lung adenocarcinoma is not one disease. Skoulidis et al. Cancer Discovery 2018 ¹⁰ described three transcriptionally distinct subtypes defined by co-mutation status: KRAS-only (K), KRAS + TP53 (KP), and KRAS + STK11/LKB1 (KL). The KL subgroup showed an objective response rate to PD-1 blockade of 7.4 percent versus 35.7 percent in the KP subgroup and 28.6 percent in K-only — a roughly five-fold difference within a single oncogene-defined population. Arbour et al. Clin Cancer Res 2018 ¹¹ extended the picture for KEAP1 co-mutation, which carried independent prognostic weight in KRAS-mutant NSCLC and clustered with worse outcomes on standard chemotherapy.

The mechanistic basis has since become measurable. Qian et al. Cancer Cell 2023 ¹² traced LKB1-deficient lung adenocarcinoma to enhanced lactate secretion via MCT4, M2 macrophage polarisation, and hypofunctional T cells; MCT4 knockout reversed PD-1 blockade resistance in syngeneic models. The KEAP1 axis runs through NRF2-mediated antioxidant reprogramming and glutamine dependency, a second metabolic shunt that intersects Branch 2. Skoulidis et al. Nat Med 2025 ¹³ is the substrate's anchor for the modern biomarker subgroup analysis — co-mutation status now reads as a clinical stratifier, not a biological footnote.

The therapeutic implication is direct. "KRAS-mutant NSCLC" as a trial-design unit dissolves on the dimension that matters most for any IO combination: T-cell exclusion. A combination programme that enrolls without co-mutation stratification is enrolling three different diseases.

PMID anchors: 29773717 · 29089357 · 37327788 · 40437272

Pancreatic ductal adenocarcinoma is mechanically anomalous among solid tumours. Hingorani et al. Cancer Cell 2003 ¹⁴ established the founding KRAS-G12D mouse model that recapitulates PanIN-to-invasive progression and produces the characteristic desmoplastic stroma — a fibroinflammatory matrix dominated by hyaluronic acid, type I collagen, and activated pancreatic stellate cells. Provenzano et al. Cancer Cell 2012 ¹⁵ quantified the consequence: PDAC tumours generate interstitial fluid pressures that exceed prior measurements for any solid tumour, collapse the microvasculature, and impose physical barriers to small-molecule perfusion. Olive et al. Science 2009 ¹⁶ made the same point through different pharmacology: Hedgehog inhibition transiently increased intratumoral vascular density and gemcitabine concentration in mice.

The clinical record is consistent and discouraging. The HALO 109-301 Phase 3 of PEGPH20 plus gemcitabine + nab-paclitaxel in hyaluronan-high 1L PDAC reported HR 1.00, mOS 11.2 versus 11.5 months (Van Cutsem et al.)¹⁷ — a clean miss against the stromal hypothesis. Vismodegib + gemcitabine (Catenacci)¹⁸ and IPI-926 + gemcitabine (Ko terminated for harm)¹⁹ closed the Hedgehog approach. Three stromal-targeting trials read negative; the field then asked not whether stromal barriers matter, but whether pharmacology arriving from outside the tumour cell can remediate them.

This is the structural reason the pathway-directed PDAC graveyard reads the way it does. The Wolpin et al. NEJM 2026 ²⁰ daraxonrasib Phase 1/2 in the 26-patient G12-mutant 300 mg 2L subgroup reported median OS 13.1 months and ORR 35 percent — single-arm data, the first peer-reviewed PDAC signal from a direct active-state RAS inhibitor that engages the tumour cell rather than depending on stromal modulation.

PMID anchors: 14706336 · 22439937 · 19460966 · 32706635 · 26527777 · 26390428 · 42090791

The KRAS mutational spectrum is non-uniform across tissues, and the asymmetry is not subtle. Prior et al. Cancer Research 2020 ²¹ systematically catalogued allele frequencies across public databases: G12C is enriched in lung adenocarcinoma (roughly 13 percent of LUAD; the dominant KRAS allele in that disease), while G12D dominates pancreatic ductal adenocarcinoma (roughly 40 percent of PDAC) and is the most common KRAS allele in colorectal cancer. HRAS mutations cluster in head and neck squamous and bladder cancers; NRAS dominates melanoma. The isoform and codon distributions have been stable across cohorts for decades and are unlikely to be sampling artefacts.

Two competing explanations sit in the literature. The first is mutagen exposure: Alexandrov et al. Nature 2013 ²² catalogued more than twenty distinct mutational signatures across 7,042 cancer genomes, including a tobacco-associated signature in lung adenocarcinoma dominated by C-to-A transversions — the substitution class that converts KRAS codon 12 GGT to TGT (G12C) preferentially. The published explanation associates lung-G12C enrichment with tobacco mutagenesis, consistent across the major LUAD cohorts. The second is tissue-specific selection: a given allele may exert different downstream signaling intensity in different epithelial contexts, with G12D producing the GTP-dwell-time profile that PDAC ductal precursors require and G12C producing the profile that LUAD type II pneumocytes tolerate. The two explanations are not mutually exclusive and the literature has not yet adjudicated cleanly between them.

The pattern dictates portfolio shape. Allele-selective inhibitors map onto the tissues where their allele concentrates; pan-RAS coverage is the strategy that decouples portfolio breadth from the allele × tissue matrix.

PMID anchors: 32209560 · 25323927 · 23945592

The G12C-OFF class produced the first systematic clinical resistance landscape for direct KRAS inhibition. Awad et al. NEJM 2021 ²³ characterised acquired resistance in adagrasib-treated patients across 17 of 38 evaluable cases: secondary KRAS mutations (Y96D/C, H95, R68; G12D/V/R; Q61H), G12C allele amplification, NRAS/HRAS/MRAS bypass, MAP2K1 activation, MET amplification, PTEN loss, oncogenic fusions, histologic transformation. Tanaka et al. Cancer Discovery 2021 ²⁴ reported the first Y96D switch-II clinical resistance to a G12C-OFF agent and showed that a tri-complex chemotype (RM-018) retained in vitro activity against the same mutation by binding outside the switch-II pocket. These routes converge on a common architecture: most G12C-OFF resistance involves reactivation of the GTP-loaded state. Xue et al. Nature 2020 ² established that RAS-dependent RTK feedback through ERBB2/3 is near-universal in the short term.

Sang et al. Cell 2026 ²⁵ now provides the first peer-reviewed clinical resistance profile for a tri-complex inhibitor: 40 daraxonrasib-treated patients across mixed tumour types (17 NSCLC, 15 CRC, 4 melanoma, 4 other), 18 of 40 with acquired alterations at progression. The dominant clinical routes are RAS Y64 binding-interface mutations (~5 percent; disrupt π-π stacking with the daraxonrasib indole) and kinase-dead BRAF (~12.5 percent; drives RAF dimerisation that attenuates CYPA recruitment to active-state RAS). Preclinical-only routes — RAS Y71 and CYPA/PPIA interface mutations — are documented in cell lines and base-editing screens but not in patient samples (CYPA is absent from clinical NGS panels, which Sang et al. flag explicitly). The detection rate of 45 percent (18/40) is identical to the G12C-OFF comparator rate of 45 percent (17/38) in Awad NEJM 2021 — same rate, different routes.

The durability question is now empirical. A positive readout that holds across post-progression follow-up answers it one way; one that collapses answers it the other — resistance converging faster than the survival signal.

PMID anchors: 34161704 · 33824136 · 31915379 · 42092352

Indication

2L+ metastatic PDAC, KRAS G12-mutant

Mechanism

KRAS-GTP · cyclophilin A · GAP-mimetic transition state

Venue

ASCO↗ 2026 Plenary LBA5, Hall B1, 3:21 PM CDT

What the literature implies

Wolpin NEJM 2026 (single-arm Phase 1/2, n=26 G12 300 mg 2L PDAC subgroup against a pooled any-RAS 2L n=38 baseline)²⁰ reports 13.1-month median OS, 35% ORR, and 8.5-month mPFS. The mPFS figure sits above the peer-reviewed G12C-OFF NSCLC mPFS comparator (KRYSTAL-1 6.5 mo, CodeBreaK 100 6.8 mo). The published Phase 3 record in mPDAC places the historical second-line floor at ~6.1 months mOS (NAPOLI-1 NAL-IRI + 5-FU/LV, Wang-Gillam Lancet 2016)²⁶. Three independent academic groups have replicated the mechanism — daraxonrasib engages the GTP-bound active-state effector lobe via cyclophilin A: Cuevas-Navarro Nature 2025 (MSKCC, Lito laboratory — the same lab as Sang et al. Cell 2026, with disclosed RVMD↗ institutional grants); Pfaff & Shokat Biol Chem 2026 (UCSF, wholly independent); Alexander JBC 2025 (NCI/Frederick, wholly independent).

Historical comparator

Twenty-five years of pathway-directed Phase 3 development in mPDAC have produced one statistically positive trial across roughly fourteen attempts: Moore PA.3 (JCO 2007, ΔmOS 0.33 mo)²⁷ is the lone positive, against a graveyard that includes Fuchs GAMMA ganitumab (Ann Oncol 2015)²⁸, Van Cutsem HALO 109-301 PEGPH20 ¹⁷, Chung S1115 selumetinib + MK-2206 ²⁹, Infante trametinib + gemcitabine ¹, Catenacci vismodegib ¹⁸, Golan POLO olaparib ³⁰, Wainberg NAPOLI-3 ³¹, and Hurwitz JANUS-1/2 ruxolitinib ³².

Unresolved questions

• ·Whether the published single-arm Phase 1/2 OS signal reproduces under randomization.
• ·How the Phase 3 ITT cohort distributes across G12 sub-alleles (G12D, G12V, G12C, G12R).
• ·Whether discontinuation due to TRAE falls below the ~15% range or above.
• ·Whether QoL on the active arm holds against investigator-choice chemotherapy.
• ·Whether the cyclophilin A safety axis (CYPA / NFAT / calcineurin) generates a clinical signal — the published literature has not measured this.

PMID anchors: 42090791 · 37590355 · 38589574 · 39476862 · 42030031 · 40473215 · 42092352 · 17452677 · 26615328 · 34096690 · 35658005 · Per-ticker report →

Indication

Multiple — NSCLC, CRC, biliary tract, advanced solid tumors

Mechanism

KRAS G12D · switch-II / Helix-3 pocket (covalent and noncovalent)

Venue

ASCO↗ 2026 Developmental Therapeutics oral session, abstracts 3006 / 3007 / 3008 (GFH375 / VS-7375 confirmed at 3008 per canonical asco_presentation_map; the two adjacent assets per advance program disclosures, pending final ASCO planner cross-check)

What the literature implies

Park NEJM 2026 (setidegrasib / ASP3082 Phase 1)³³ reports PDAC G12D ORR 24%, mPFS 3.0 mo, mOS 10.3 mo at peer-review tier. Zhou Cancer Cell 2024 ³⁴ covers HRS-4642 preclinically with a one-sentence early-clinical mention. Hallin Nature Medicine 2022 ³⁵ reports MRTX1133 preclinical activity; the MRTX1133 Phase 1 (NCT05737706↗) terminated without a peer-reviewed clinical readout.

Historical comparator

Peer-reviewed clinical evidence for direct G12D engagement is sparse. Setidegrasib reads 24% ORR in the PDAC G12D subgroup (n=21) and 36% PR rate in the NSCLC G12D subgroup (n=45) per Park NEJM 2026 ³³. Both subgroups sit at or below the published G12C-OFF NSCLC comparator band of 30–50% (CodeBreaK 100 37.1%, KRYSTAL-1 42.9%, divarasib 53.4%).

Unresolved questions

• ·Whether allele-selective G12D engagement extends meaningfully to CRC and cholangiocarcinoma at peer-review tier.
• ·Whether the three agents differ structurally in switch-II pocket engagement vs RAS-effector binding.
• ·Whether resistance follows the switch-II covalent pattern or generates new escape geometries.
• ·How the G12D allele frequency across indications (PDAC ~40%, NSCLC ~30% of G12) interacts with allele-selective agent geographic strategy.

PMID anchors: 41879829 · 38942026 · 36216931 · Per-ticker report →

Indication

NSCLC G12D, Phase 1/2

Mechanism

KRAS G12D · switch-II / Helix-3 · oral PK

Venue

ASCO↗ 2026 Lung Cancer Metastatic rapid oral, abstract 8516

What the literature implies

First KRAS G12D inhibitor in a dedicated NSCLC oral session at ASCO↗ 2026. No prior peer-reviewed Phase 1/2 efficacy publication identified at search date. AACR↗ 2026 Plenary disclosure (RVMD↗ IR, 2026-04-19) reports zoldonrasib NSCLC G12D ORR 52% (95% CI 32–71, n=27) at the conference-frontier tier; TSN1611 lands in the same indication space but a different mechanism class (switch-II vs active-state).

Historical comparator

Active-state RAS in NSCLC G12D (zoldonrasib) is currently at conference-frontier T4 tier. TSN1611 is the first switch-II G12D NSCLC asset to occupy a dedicated oral slot.

Unresolved questions

• ·Whether TSN1611 ORR clears the bar set by adagrasib KRYSTAL-1 in G12C NSCLC (42.9% ORR, mPFS 6.5 mo at peer-review tier).
• ·Whether oral PK delivers the exposure-response curve required for durable inhibition.
• ·Whether resistance emerges through the switch-II Y96-class mutations that defined the G12C-OFF resistance landscape (Tanaka 2021)²⁴.

PMID anchors: 33824136 · 35658005 · Per-ticker report →

Indication

Advanced or metastatic PDAC, Phase 2a

Mechanism

MEK1/2 · downstream of KRAS · MAPK reactivation feedback

Venue

ASCO↗ 2026 GI Pancreatic rapid oral, abstract 4013

What the literature implies

Three peer-reviewed MEK trials in PDAC have read negative overall. Trametinib + gemcitabine (Infante)¹ reports HR 0.98 with no signal. Selumetinib + MK-2206 S1115 (Chung) reads HR 1.37, with the control arm outperforming the experimental. Atebimetinib must clear that published MEK + chemotherapy record as its historical comparator.

Historical comparator

Pathway-directed MEK + chemotherapy in mPDAC has produced no Phase 3 OS win in the published record.

Unresolved questions

• ·Whether atebimetinib achieves a meaningful PFS or OS signal vs the gem/nab-pac historical control.
• ·Whether the MEK + chemo backbone shows differential activity in KRAS subtype subgroups (G12D vs G12V vs G12R).
• ·Whether the RASolute 303 1L MEK + chemo comparator framework is informed by these data.

PMID anchors: 24915778 · Per-ticker report →

Indication

HLA*A2(-) metastatic uveal melanoma

Mechanism

PKC + c-MET · uveal melanoma signaling

Venue

ASCO↗ 2026 Melanoma/Skin Cancers oral session LBA9503, S100bc, June 1, 8–11 AM CDT

What the literature implies

OptimUM-02 has reported HR 0.42 and mPFS 6.9 vs 3.1 mo per IDEAYA↗ disclosure. NDA H2 2026 per sponsor guidance. Peer-reviewed publication anticipated 2026.

Historical comparator

Metastatic uveal melanoma reads a historically poor survival record; HLA*A2(-) patients fall outside the tebentafusp-eligible population.

Unresolved questions

• ·Whether subgroup HRs hold across baseline LDH and tumor burden strata.
• ·Whether the OS curve diverges early or late — relevant to NDA framing.
• ·Whether the combination tolerability profile supports a chronic-dosing regimen.

Per-ticker report →

Indication

Desmoid tumor

Mechanism

Notch / gamma-secretase · CTNNB1 wnt pathway

Venue

ASCO↗ 2026 Sarcoma oral abstract session, S100bc, May 30, 3–6 PM CDT, varegacestat oral (pending final ASCO planner cross-check)

What the literature implies

Phase 3 HR 0.16 per IMNM↗ disclosure. NDA filed 2026-04-29 per sponsor guidance. ASCO↗ oral fills in subgroup detail across CTNNB1 mutation strata.

Historical comparator

Desmoid tumor progresses heterogeneously, with periods of spontaneous regression; historical control PFS varies widely.

Unresolved questions

• ·Whether ORR sustains above 50% across CTNNB1 subgroups (S45F, T41A, S45P).
• ·Whether the durability of response extends to two years of treatment.
• ·Whether the safety profile supports indefinite dosing in a chronic disease.

Per-ticker report →

Indication

2L head and neck + cervical cancer

Mechanism

nectin-4 · ADC payload delivery

Venue

ASCO↗ 2026 DT oral

What the literature implies

CRB-701 ORR reported at 47.6% in 2L head and neck per CRBP↗ disclosure. Zero peer-reviewed efficacy publications identified at search date.

Historical comparator

Enfortumab vedotin anchors the nectin-4 ADC class in urothelial cancer at peer-review tier. Head and neck and cervical extensions sit at conference-frontier tier.

Unresolved questions

• ·Whether the 47.6% ORR replicates under independent imaging review.
• ·Whether the safety profile in head and neck mirrors urothelial enfortumab signals or shifts.
• ·Whether nectin-4 expression heterogeneity in head and neck explains responder vs non-responder patterns.

Per-ticker report →

Indication

Alpha-1 antitrypsin deficiency

Mechanism

GalNAc-conjugated RNA editing · liver delivery

Venue

Already disclosed pre-meeting; ASCO↗ follow-up presentation

What the literature implies

WVE↗ GalNAc cleared 11 μM PiZ-AAT where one LNP did not, per sponsor disclosure. Beam Therapeutics LNP comparator cleared the same target harder. Peer-reviewed comparative data have not yet been published.

Historical comparator

The RNA-editing class for AATD divides on the delivery-route choice — LNP versus GalNAc; the published mechanism literature has not yet resolved that comparison.

Unresolved questions

• ·Whether durability of AAT correction extends beyond the disclosed timepoint.
• ·Whether off-target editing risk differs by delivery-route choice.
• ·Whether the published comparative claim (WVE↗ GalNAc vs Beam LNP) reproduces in larger cohorts.

Per-ticker report →