Why KRAS-mutant PDAC needs more than RAS inhibition alone

More than 90% of pancreatic ductal adenocarcinomas are driven by an activating mutation in KRAS, most often at codon 12 (G12D, G12V, G12R)³. For two decades this made the central oncogenic driver effectively undruggable in PDAC, because the common pancreatic mutations are not the G12C allele that covalent OFF-state inhibitors such as sotorasib and adagrasib were built for.

Pan-RAS(ON) inhibitors changed that calculus. But monotherapy against active RAS is not curative: in the phase 1/2 RMC-6236-001 trial, daraxonrasib produced objective responses in roughly a quarter to a third of previously treated RAS-mutant PDAC patients, a median progression-free survival of about 8.5 months and a median overall survival of about 13.1 months — meaningful against historical second-line chemotherapy, yet typically followed by relapse⁴. Preclinically, RMC-7977 drives deep regressions across PDAC models that are later undermined by adaptive and acquired resistance^1,2. The question is no longer whether to hit RAS, but what to hit alongside it.

How daraxonrasib (RMC-6236) and RMC-7977 block active RAS-GTP

Daraxonrasib and its preclinical analogue RMC-7977 are reversible tri-complex inhibitors. Each compound binds the abundant intracellular chaperone cyclophilin A, and the resulting binary complex acquires high affinity for the active, GTP-bound state of KRAS, NRAS and HRAS — mutant and wild-type alike^1,2. The neomorphic surface created at the cyclophilin A–RAS interface sterically occludes the effector-binding region, so RAS can no longer recruit RAF or PI3K, and both the MAPK and PI3K/AKT arms downstream of RAS are suppressed.

The "multi-selective" design matters for resistance. Allele-selective inhibitors leave wild-type RAS isoforms free to reactivate ERK through receptor-tyrosine-kinase feedback, producing the rapid pERK rebound that limits durability. By suppressing both mutant and wild-type RAS-GTP simultaneously, pan-RAS(ON) agents blunt that rebound^2,9. In PDAC, RMC-7977 also produces a tumour-selective response: treated tumours show waves of apoptosis and sustained proliferative arrest while normal tissues experience only transient slowing of proliferation².

ROCK2 in the pancreatic actin cytoskeleton, EMT, and chemoresistance

ROCK2 is a serine/threonine kinase and a principal effector of the Rho GTPases. It phosphorylates the myosin phosphatase targeting subunit MYPT1 (Thr696/Thr850), increasing actomyosin contractility, stress-fibre assembly and focal adhesion maturation⁶. In pancreatic cancer, ROCK1 and ROCK2 expression rises with tumour stage and grade, and ROCK genomic amplification correlates with poorer survival⁶. The RhoA/ROCK2 axis supports the invasive, mesenchymal phenotype that makes PDAC so metastatic.

Inhibiting ROCK reverses part of that programme. Knockdown or pharmacological inhibition of ROCK2 partially restores an epithelial state — downregulating mesenchymal markers such as N-cadherin and vimentin while raising E-cadherin — and reduces motility⁵. ROCK2 also props up a ZEB1-dependent DNA-damage-repair axis: in gemcitabine-resistant pancreatic cells, ROCK2 upregulates ZEB1, and silencing ROCK2 re-sensitises those cells to gemcitabine in a ZEB1-dependent manner⁵. Separately, the dual ROCK1/2 inhibitor AT13148 blunts PDAC invasion and has been used to remodel the stiff, desmoplastic stroma and prime tumours for chemotherapy⁶.

Focal-adhesion adaptive resistance: the cytoskeletal bypass to RAS inhibition

The most direct argument for the combination comes from how PDAC cells behave when KRAS is switched off. In KRAS-inhibited pancreatic cells, surviving cells enter a reversible, non-genetic and non-transcriptional state characterised by prominent focal adhesion plaques and increased tyrosine phosphorylation of adhesion proteins — Src, focal adhesion kinase (FAK), tensin, paxillin, talin and vinculin⁷. These cells adhere faster, build larger adhesion structures, and become more dependent on attachment for viability.

That attachment dependency is the vulnerability. If the cytoskeletal and adhesion machinery sustaining the Kras-inhibited state is disrupted, the cells should be pushed toward anoikis. ROCK2 sits at the heart of that machinery, governing the contractility and adhesion turnover the persister state relies on. The same logic explains why FAK inhibition reprogrammes the fibrotic, immunosuppressive PDAC stroma and sensitises tumours to checkpoint immunotherapy⁸ — adhesion signaling is a shared node between drug tolerance and the microenvironment.

The mechanistic rationale for combining pan-RAS(ON) and ROCK2 inhibition

The two agent classes act on orthogonal, non-redundant layers of PDAC biology. Daraxonrasib removes the oncogenic signal; ROCK2 inhibition disables the cytoskeletal and adhesion programme that PDAC cells deploy to survive when that signal is gone. In principle, ROCK2 inhibition could collapse the focal-adhesion-dependent persister state, suppress the EMT and ZEB1 programmes that fuel chemoresistance, and soften the stroma that limits drug delivery and immune access — while pan-RAS(ON) inhibition prevents the pERK rebound that allele-selective drugs cannot.

What the RAS plus ROCK2 evidence shows in PDAC — an insufficient-evidence verdict

Here the honest part of the analysis matters more than the hypothesis. Across three structured BioSkepsis searches — one on the molecular pathways triggered by the combination, one on RMC-6236/ROCK synergy in KRAS- or NRAS-mutant models, and one on the combined effect on the actin cytoskeleton and adaptive resistance in PDAC — the platform found extensive independent characterisation of each agent class and zero studies evaluating them together. Rather than stitch the two literatures into a synthetic "synergy," BioSkepsis returned an explicit insufficient-evidence response and proposed concrete next searches.

This is the difference between citation-grounded reasoning and pattern-completion. A general-purpose model asked the same question will often produce a confident, fluent account of a daraxonrasib-plus-ROCK2 mechanism that reads as established fact; BioSkepsis instead declines to claim what the corpus does not support and shows the gap. The combination is biologically plausible and worth testing. It is not yet a result. For more on how this evidence posture compares with autonomous hypothesis systems, see BioSkepsis vs Co-Scientist and the Research Hub.

Who should use this analysis in pancreatic cancer research

Frequently asked questions