Multi-target, ensemble-based virtual screening yields novel allosteric KRAS inhibitors at high success rate.

RAS mutations account for >15% of all human tumors, and of these ~85% are due to mutations in a particular RAS gene: KRAS. Recent studies revealed that KRAS harbors four druggable allosteric sites. Here, we have (a) used molecular simulations to generate ensembles of wild type and four major oncogenic KRAS mutants (G12V, G12D, G13D, and Q61H); (b) characterized the druggability of each allosteric pocket in each protein; (c) conducted extensive ensemble-based virtual screening using pocket-tailored ligand libraries; (d) prioritized hits through hierarchical postdocking analysis; and (e) validated predicted hits with NMR. Of the 785 diverse potential hits identified by our in silico analysis, we tested 90 for their ability to bind KRAS using NMR and found that nine cause backbone amide chemical shift perturbations of residues near the functionally responsive switch loops, suggesting potential binding. We conducted detailed biophysical analyses on a novel indole-based compound to demonstrate the potential of our workflow to yield lead compounds. We believe the detailed information documented in this work regarding the druggability profile of each allosteric site and the chemical fingerprints of compounds that target them will serve as vital resources for future structure-based drug design efforts against KRAS, a high-value target for cancer therapy.