KRAS G12D — drugging the bigger cousin in pancreatic cancer

The G12C playbook does not transfer to G12D. The whole reason sotorasib and adagrasib work is the engineered cysteine at position 12 — a covalent handle that does not exist in any other KRAS variant. G12D is a bigger problem in two senses: it is the most common KRAS mutation overall, dominating pancreatic ductal adenocarcinoma (PDAC), and it forces the chemistry to find a different way in. The field has spent the past four years building that alternative, and the 2026 clinical readouts are the first meaningful evidence it might work.

Why G12D is harder than G12C

Mutations at codon 12 are the bulk of oncogenic KRAS. G12C accounts for roughly 13% of KRAS-mutant tumors and is concentrated in NSCLC. G12D accounts for closer to 36% of KRAS-mutant tumors and is the dominant variant in PDAC, where it appears in around half of all cases. It also shows up in colorectal and endometrial cancers, so a working G12D drug is a much bigger therapeutic market than G12C was.

The structural problem is the residue itself. Aspartate is negatively charged at physiological pH and offers no nucleophile to react with. Every covalent G12C inhibitor — sotorasib, adagrasib, divarasib, garsorasib — uses an acrylamide warhead that forms a Michael adduct with the Cys12 thiol. Bolt that same warhead onto a G12D-shaped binder and there is nothing for it to bond to. The two viable strategies are non-covalent binders that get enough affinity from shape and electrostatics alone, or bifunctional approaches (tri-complex inhibitors, degraders) that do not rely on a single hook on the target.

MRTX1133: the high-affinity non-covalent path

MRTX1133 (Mirati Therapeutics, now Bristol-Myers Squibb) was the first credible KRAS G12D-selective small molecule. The discovery paper from Wang et al. (2022) reports a biochemical IC50 below 2 nM against KRASG12D-GDP with roughly 700-fold selectivity over wild-type KRAS, and an in-cell pERK IC50 of about 5 nM in G12D-mutant lines vs. greater than 1000-fold weaker in WT-KRAS-driven lines. Hallin et al. (2022) followed up with the in vivo case: tumor regression in 8 of 11 PDAC patient-derived xenograft models. The selectivity comes from a key salt bridge with the mutant D12 carboxylate — exactly the kind of pose you can rationalize from a docking run, since the wild-type protein has no charged residue there to engage.

The MRTX1133 problem is route. The original compound is not orally bioavailable and the Phase 1/2 trial dosed it intravenously. An oral prodrug program is reportedly in progress (Lin et al., ACS Omega 2023), but as of mid-2026 the leading clinical-stage G12D drug in oral dosing is coming from a different chemotype entirely.

RAS(ON) tri-complex inhibitors: Revolution Medicines

Revolution Medicines built a fundamentally different mechanism. Instead of competing with effectors for the RAS surface, the RAS(ON) class glues RAS-GTP to cyclophilin A (CypA) through a small-molecule bridge. The resulting ternary complex sterically occludes the effector binding site — RAS is still GTP-loaded and conformationally “on”, but it cannot signal because RAF and PI3K cannot reach it. Holderfield et al. (Nature 2024) characterized the mechanism for the tool compound RMC-7977 and the clinical candidate RMC-6236.

• Daraxonrasib (RMC-6236) — a multi-selective RAS(ON) inhibitor that hits G12D, G12V, G12R, G13D, and wild-type RAS-GTP. The Phase 1 RMC-6236-001 trial reported a 47% objective response rate in previously untreated metastatic PDAC with RAS-mutant disease. FDA granted breakthrough therapy designation for previously treated metastatic PDAC harboring KRAS G12X mutations. The Phase 3 RASolute 302 study is reading out across 2026.
• Zoldonrasib (RMC-9805) — a G12D-selective covalent RAS(ON) inhibitor. The covalent handle here is not a cysteine; the chemistry engages the mutant aspartate itself. Early Phase 1 signals in G12D PDAC were reported at ESMO 2025.

The mechanistic appeal is that the tri-complex hits RAS in its active state, not the GDP-bound state that the G12C drugs wait for. That removes a kinetic ceiling — you do not need to outpace the nucleotide cycle to engage the target.

What resistance probably looks like

Too early for clinical resistance data on G12D the way we have for G12C, but two patterns are predictable from first principles. Switch-II pocket mutations (Y96, H95, R68) that broke sotorasib will likely break MRTX1133 — it binds in the same general location. The tri-complex compounds are vulnerable to a different class: cyclophilin A loss-of-function, alterations in the RAS surface where CypA is glued in, or upregulation of RAS isoforms outside the inhibitor’s coverage. The first cell-line resistance screens against RMC-6236 published in 2025 (Bekele et al., not yet peer-reviewed) saw a different escape spectrum than the G12C program — closer to bypass via PI3K and RTK reactivation than on-target second-site mutation.

Try the docking yourself

The MRTX1133 binding pose is captured in PDB 7T47 — switch-II pocket opened in the G12D mutant, with the inhibitor making the key salt bridge to D12. Open Studio and pick KRAS from the target catalog with G12D from the mutation chips. Liganx will dock against the G12D receptor and you can run the same ligand against G12C and wild-type in parallel — the ΔΔ between G12D and WT for MRTX1133 should land around 2 kcal/mol in Vina, which is the docking-level signature of that salt-bridge selectivity story. The ADMET panel will flag oral bioavailability concerns for high-MW analogs, which is most of the chemistry challenge in this series.

Liganx is molecular docking online and free in the browser, set up for exactly this kind of mutation-vs-wildtype comparison. If you want to run molecular docking on KRAS G12D without a local install, it is the fastest path.

Primary sources

• Wang X, et al. Identification of MRTX1133, a Noncovalent, Potent, and Selective KRAS G12D Inhibitor. J Med Chem 65, 3123-3133 (2022). doi:10.1021/acs.jmedchem.1c01688
• Hallin J, et al. Anti-tumor efficacy of a potent and selective non-covalent KRASG12D inhibitor. Nat Med 28, 2171-2182 (2022). doi:10.1038/s41591-022-02007-7
• Holderfield M, et al. Concurrent inhibition of oncogenic and wild-type RAS-GTP for cancer therapy. Nature 629, 919-926 (2024). doi:10.1038/s41586-024-07205-6
• Jiang J, et al. Translational and Therapeutic Evaluation of RAS-GTP Inhibition by RMC-6236 in RAS-Driven Cancers. Cancer Discov 14, 994-1017 (2024). doi:10.1158/2159-8290.CD-24-0027