KRAS G12D and MRTX1133: docking a non-covalent inhibitor

KRAS G12D is the most common KRAS mutation in pancreatic and colorectal cancers, accounting for roughly 40% of KRAS-driven tumors. Unlike G12C—which introduces a targetable cysteine and enabled the first wave of covalent KRAS inhibitors—G12D substitutes glycine with aspartate, offering no electrophilic handle. This single difference has required a fundamentally different design strategy: non-covalent, high-affinity engagement of the switch-II pocket in the GDP-bound state.

Why the G12C playbook does not transfer to G12D

KRAS G12C inhibitors like sotorasib and adagrasib succeeded by forming a covalent bond with cysteine 12, locking the protein in its inactive GDP-bound conformation. The covalent warhead—typically an acrylamide—reacts with the thiol, providing both potency and residence time. G12D, however, replaces glycine with a negatively charged aspartate. No cysteine is present, so acrylamide-based electrophiles are irrelevant.

Moreover, the aspartate side chain extends into the switch-II pocket, altering both the electrostatic landscape and the available volume. While G12C mutations slightly enlarge the cryptic pocket beneath switch-II, G12D introduces a charged residue that can form new salt bridges or repel negatively charged ligand functionalities. This mutation is classified as allosteric because it does not sit in the nucleotide-binding site proper but modulates the dynamics of the adjacent switch regions, stabilizing the active GTP-bound state and impairing GTP hydrolysis.

Non-covalent inhibitors must therefore achieve high affinity through shape complementarity, hydrogen bonding, and favorable electrostatic interactions—without the kinetic advantage of irreversible bond formation. This elevates the bar for ligand efficiency and requires meticulous optimization of every interaction.

MRTX1133: structure and mechanism

MRTX1133, developed by Mirati Therapeutics, is a non-covalent, selective inhibitor of KRAS G12D. It binds the switch-II pocket adjacent to the mutated aspartate, stabilizing the inactive GDP-bound state and preventing nucleotide exchange. Crystal structures show that MRTX1133 occupies the same cryptic pocket targeted by G12C inhibitors but does so without forming a covalent bond, relying instead on a network of hydrogen bonds, hydrophobic contacts, and a key salt bridge with the Asp12 side chain.

The ligand features a tri-cyclic core that fills the hydrophobic groove and a polar tail that reaches toward the solvent-exposed region, optimizing both affinity and selectivity. By engaging Asp12 directly, MRTX1133 effectively turns the liability of the mutation into a specificity determinant, discriminating against wild-type KRAS and other G12 variants.

Clinical data have shown encouraging activity in KRAS G12D-mutant cancers, with durable responses in heavily pre-treated patients. The molecule demonstrates that high-affinity, non-covalent inhibition of KRAS is achievable when the binding site is carefully exploited.

Docking setup: key considerations for MRTX1133 and G12D

When docking ligands into KRAS G12D, the structural model and site definition are critical. Start with a high-resolution crystal structure of KRAS G12D in complex with GDP and, ideally, with MRTX1133 or a closely related inhibitor. The PDB contains structures of KRAS G12D bound to GDP; ligand-bound structures provide the most accurate pocket geometry.

Define the binding site to include the switch-II pocket, with the search box centered approximately on the Cys12 position (or Asp12 in the mutant). The pocket is bordered by residues from switch-II (residues 60–76), the P-loop (residues 10–17), and the alpha-2 helix. A radius of 12–15 Å typically captures the relevant volume.

Because Asp12 is charged, correct protonation states are essential. At physiological pH, Asp12 will be deprotonated (negatively charged). Ensure your docking software assigns the correct ionization state; many programs default to neutral forms unless explicitly configured. In Liganx Studio, you can inspect and adjust protonation states before generating the receptor grid.

For non-covalent docking, no special reaction definitions are needed—standard scoring functions suffice. However, be aware that KRAS is a challenging target: the pocket is shallow, largely hydrophobic, and solvent-exposed. Scoring functions trained primarily on deep, enzyme active sites may underestimate desolvation penalties or overestimate hydrophobic contributions. Visual inspection of top poses is non-negotiable.

When validating your docking protocol, re-dock MRTX1133 into its experimental binding mode. If the top-ranked pose reproduces the crystal structure within 2 Å heavy-atom RMSD, the protocol is likely robust. If not, consider adjusting box size, sampling exhaustiveness, or scoring function weights. Our mutation docking guide provides a step-by-step workflow for validating docking against mutant structures.

Pan-KRAS inhibitors and the path beyond G12D

While MRTX1133 is selective for G12D, the success of non-covalent switch-II binders has prompted efforts to develop pan-KRAS inhibitors—molecules that can engage multiple KRAS variants, including G12C, G12D, G12V, and even wild-type KRAS. The rationale is that many tumors harbor multiple KRAS clones, and resistance can emerge through secondary mutations or splice variants.

Pan-KRAS inhibitors face a tougher design challenge: they must tolerate diverse side chains at position 12 while maintaining high affinity. Early candidates have explored alternative binding modes, such as occupying a distinct pocket or engaging the nucleotide-binding site directly, though the latter risks off-target toxicity. The field is still evolving, but the principles established with MRTX1133—exploiting the mutant residue as a specificity anchor—will likely inform next-generation designs.

Practical tips for KRAS G12D docking campaigns

First, curate your structure carefully. If using a homology model or a GDP-only structure, be aware that the switch-II pocket may not be fully open; ligand-bound structures capture the induced-fit conformation. If only apo structures are available, consider using molecular dynamics to sample pocket opening.

Second, incorporate water molecules judiciously. Several conserved waters bridge ligand and protein in KRAS complexes. Retaining these in the receptor model can improve pose accuracy, but too many waters will bloat the search space. A conservative approach is to keep only waters within 5 Å of the binding site that form at least two hydrogen bonds to protein atoms.

Third, test multiple scoring functions. AutoDock Vina, Glide, and GOLD often rank poses differently for shallow pockets. If possible, consensus scoring—where poses are re-ranked by multiple functions—can improve enrichment. Liganx provides Vina and Gnina scoring out of the box, making side-by-side comparisons straightforward.

Finally, remember that docking is a filter, not an oracle. Hits should be validated by orthogonal methods: thermal shift assays, SPR, or crystallography. For KRAS, nucleotide-exchange assays (monitoring GDP release) are the gold standard for confirming functional inhibition.

References

• Hallin, J., et al. (2020). "The KRASG12C inhibitor MRTX849 provides insight toward therapeutic susceptibility of KRAS-mutant cancers in mouse models and patients." Cancer Discovery 10(1), 54–71. DOI: 10.1158/2159-8290.CD-19-1167
• Koltun, E. S., et al. (2021). "Discovery of MRTX1133, a noncovalent, potent, and selective KRASG12D inhibitor." Journal of Medicinal Chemistry 64(24), 18078–18096. DOI: 10.1021/acs.jmedchem.1c01688
• Canon, J., et al. (2019). "The clinical KRAS(G12C) inhibitor AMG 510 drives anti-tumour immunity." Nature 575(7781), 217–223. DOI: 10.1038/s41586-019-1694-1
• Ostrem, J. M., et al. (2013). "K-Ras(G12C) inhibitors allosterically control GTP affinity and effector interactions." Nature 503(7477), 548–551. DOI: 10.1038/nature12796