Prioritizing Candidate YB‑1 Cold Shock Domain Ligands via 5D-ElectroShape and Boltz Generative Cofolding

Y-box binding protein 1 (YB-1) acts as a key oncoprotein, with its highly conserved cold shock domain (CSD) driving core nucleic acid-binding functions in cancer progression. Despite its therapeutic relevance, the CSD is considered "undruggable" due to the lack of a well-defined hydrophobic pocket. To overcome the limitations of rigid docking in this transient pocket, here, we present a computational ligand-prioritization workflow designed to generate experimentally testable hypotheses for this challenging target. We applied the 5D-ElectroShape formalism combined with our in-house MolPrism clustering toolkit to compress a library of ∼11,000 RNA-protein interaction-focused small molecules into 70 representative diversity centroids. These candidates were subjected to ab initio complex prediction using the deep learning-based protein folding algorithm Boltz-1, which successfully modeled the induced-fit requirements of the CSD cryptic pocket. A structure-driven selection strategy, utilizing AlphaFold-derived confidence metrics (ipTM ≥ 0.88), prioritized 18 high-confidence candidate complexes for further analysis. Subsequent molecular dynamics (MD) simulations and MM/GBSA binding free-energy calculations were used to prioritize compounds that maintained stable pocket engagement under dynamic conditions. Among the prioritized compounds, F3382-0454 showed stable cryptic-pocket occupancy during MD simulations and ligand-efficiency values comparable to known YB-1 inhibitor SU056, supporting its selection as a top-ranked candidate scaffold for future testing. Taken together, these results support the use of chemical-space compression followed by generative cofolding and MD-based refinement as a practical workflow for generating hypotheses about cryptic YB-1 ligands, while experimental validation and prospective benchmarking remain necessary.