ProCV: A 3D similarity grouping method for enhanced protein pocket recognition and ligand interaction analysis

Efficient identification of protein binding pockets is critical for accurately predicting protein-ligand interactions. Traditional sequence-based methods often fail to capture structural complexity and require extensive conformational sampling, limiting both efficiency and accuracy. To overcome these challenges, we present ProCV, an innovative structure-based prediction method that utilizes advanced spatial recognition techniques-specifically, 3D similarity grouping in the Hough space-to enhance precision and speed. ProCV employs uniform spatial sampling, KD-tree structures, and the 3D Hough transform for accurate binding pocket identification. Comparative analyses on datasets from the Protein DataBank (PDB), scPDB, and BioLip demonstrate that ProCV offers high specificity and sensitivity with reduced false positives. Its similarity assessment framework accurately characterizes the spatial arrangement of 3D protein structures, facilitating precise binding site localization. These findings highlight ProCV's robustness, precision, and flexibility in identifying binding residues at atomic resolution within 3D structures, affirming its value in structural bioinformatics for protein-ligand interaction studies.