Beyond Venetoclax: A Machine-Guided Pipeline for Drugging Mutant-Specific Cryptic Pockets in Undruggable Cancer Targets

Cryptic pockets offer a new route to drugging previously undruggable cancer targets. I present a machine-guided pipeline that integrates long-timescale molecular dynamics (MD), Markov State Modeling (MSM), time-lagged independent component analysis (TICA), and hydration thermodynamics to discover and validate mutant-specific cryptic pockets. This strategy departs from traditional static docking by capturing metastable, ligandable states not visible in crystal structures. Applied across TP53, KRAS, MYC, and SRC kinase mutants, the pipeline reveals selective pocket emergence in mutants, validated via per-residue energy decomposition and PLIP-based contact analysis. I introduce the Cryptic Dynamic Druggability Score (CDDS), a state-level metric combining pocket lifetime, water displacement, and contact persistence. Together, these methods de-risk experimental assays by pinpointing when and where to probe for ligand binding. Results confirm that cryptic pockets can be computationally validated in silico with precision sufficient to guide fragment NMR, soaking, or ITC. This pipeline enables rational design against targets long considered inaccessible to small molecules