SandCat: OpenMM Framework for Prospective Cryptic Pocket Discovery

Proteins are dynamic biomolecules, yet often in drug discovery they are treated as static entities. A relevant example of this are cryptic pockets, which are pockets that are not present in the apo structure but reveal themselves upon binding to a particular ligand. Methods based on molecular dynamics (MD) simulations can dynamically probe the protein structure and have been validated on known cryptic pocket sites. SWISH-X is one of the most prominent methods in this field. Previously, it was only available with GROMACS and PLUMED, which limited its speed. Here, we present SandCat, an OpenMM-based toolkit that implements this enhanced sampling method for cryptic pocket discovery. SandCat reproduces previous results from SWISH-X with twice the computational speed. Furthermore, SandCat provides standardised protocols for prospective use on targets where cryptic pockets are not known. We tested the SandCat pipeline on challenging targets, and recovered every characterised cryptic site within the top six candidate sites, with no sites detected for a negative control. We believe that the increased speed and the implementation of a prospective toolkit will enable SandCat to be used at scale to unlock previously undruggable targets.