GalaxyCDock: Webserver for Covalent Protein-Ligand Binding Mode Prediction

Covalent ligands represent small molecules including a reactive moiety that forms a covalent bond, enabling the targeting of proteins that are otherwise difficult to modulate. Accurate binding prediction is critical for achieving target specificity and minimizing off-target effects. However, publicly available computational tools remain limited in both accessibility and accuracy. To address this gap, we developed GalaxyCDock, a web server for covalent protein-ligand docking. GalaxyCDock predicts the binding modes of covalent ligands by employing the efficient pose sampling of GalaxyDock2 and a deep learning-based scoring function, GalaxyDock-DL. GalaxyCDock outperformed existing tools (AutoDock4, DOCK6) across standard and newly curated datasets. GalaxyCDock achieved high performance in both re-docking (up to 80%) and cross-docking (up to 61%). Furthermore, GalaxyCDock efficiently serves as a practical alternative to models like AlphaFold3 and Boltz-2 when receptor structure information is available. GalaxyCDock is publicly available at https://galaxy.seoklab.org/cdock.