Search for a command to run...
The persistent failure to drug mutant forms of key oncogenes—such as TP53, KRAS, and MYC—is largely due to their lack of stable, visible binding pockets in crystal structures. These proteins are frequently labeled “undruggable,” not because they cannot be modulated, but because their ligandable sites are hidden, transient, or mutant-specific. To address this, I developed a machine-guided, biophysically grounded pipeline that dynamically reveals cryptic pockets emerging only under specific conformational and hydration conditions. This pipeline integrates long-timescale molecular dynamics (MD), Markov State Modeling (MSM), time-lagged independent component analysis (TICA), voxel-based hydration thermodynamics (ΔG_water), and ligand–residue contact persistence analysis to predict when and where a cryptic pocket forms—and whether it is physically capable of supporting ligand binding. Unlike traditional docking or e-pharmacophore workflows, my pipeline explicitly resolves conformational ensembles into metastable macrostates and ranks them by a novel Cryptic Dynamic Druggability Score (CDDS). CDDS integrates pocket lifetime, water displacement energetics, and contact persistence, providing a state-specific and interpretable druggability metric. The entire framework was applied across four challenging cancer targets: KRAS G12D, MYC Arg58Ala, TP53 Arg248Ser, and SRC kinase. In each case, I identified either (a) a cryptic pocket emerging selectively in the mutant with supportive hydration and binding energetics, or (b) a false positive that failed all dynamic criteria, despite high static druggability scores. In KRAS G12D, a side-chain gate absent in the wild-type opened to form a metastable pocket with CDDS ~0.55, ΔG_water ≈ –3.8 kcal/mol, and mutant-selective ligand binding. In MYC Arg58Ala, a cryptic site not present in the wild-type formed via side-chain deletion and surface charge inversion; a designed compound (Compound 32) bound with MM/GBSA ≈ –21 kcal/mol and persistent PLIP contacts. SRC kinase, a classical toxicity-prone ATP-site target, revealed a cryptic site outside the active site. This site was captured dynamically and bound Analog 7 with high affinity (ΔG ≈ –31 kcal/mol) without touching the ATP pocket. In contrast, TP53 Arg248Ser showed a high Fpocket static score (0.708) but failed to sustain a pocket during MD. No meaningful ligand retention, water displacement, or stable macrostate pocket volume was observed—illustrating the power of CDDS and dynamics to reject false positives. The platform also incorporates a novel pharmacophore merging strategy using privileged scaffolds from FDA-approved and 3D-rich fragment libraries, guided entirely by dynamic pocket features rather than known ligands. Collectively, this approach enables rational design against targets previously inaccessible, de-risks biophysical assays by guiding when and where to probe, and offers a reproducible computational method for fragment hit validation, NMR probing, or soaking trials. This work provides the first integrated, mutation-specific, dynamic cryptic pocket discovery framework validated across multiple targets. It represents a paradigm shift from structure-based design to time-aware, water-sensitive, functionally annotated druggability profiling. This pipeline is currently patent-pending and offers immediate translational value in precision oncology, especially for undruggable pediatric and resistant cancer mutations.