HOCUSPOCKETS: A COMPUTATIONAL TOOL FOR AUTOMATED PRIORITIZATION OF PROTEIN TARGETS BASED ON DRUGGABILITY

This work presents the development of HocusPockets (Holistic Computational Utility
for Scoring Pockets), a computational tool designed to automate the prioritization of
biological targets in microbial proteomes, based on the integrated analysis of protein
interaction site druggability. Starting from protein 3D structures in PDB format, the
program automatically runs different cavity prediction algorithms, including Fpocket
and ConCavity — the latter incorporating classical methods such as LIGSITE,
PocketFinder, and Surfnet.In addition to geometric features, HocusPockets integrates
quantum-derived descriptors based on the spatial distribution of frontier molecular
orbitals (HOMO and LUMO), commonly associated with catalytic activity in enzymes.
These descriptors allow the identification of highly reactive sites, providing
complementary information to traditional structural analysis.The program’s workflow is
divided into five main steps: (1) data input, supporting single or multiple files; (2)
reactivity analysis, aiming to detect potentially catalytic regions; (3) cavity detection
and comparison, where grid-based representations of internal protein spaces are
generated and participating residues are extracted and matched across methods; (4)
cavity scoring, based on geometric and physicochemical descriptors; and (5) output
generation, producing structured files in CSV and PDB formats. These files contain
data on each cavity, the methods by which they were identified, associated
descriptors, and the cavity’s geometric center, which can be used for downstream
applications such as molecular docking setup.As a final objective, the tool aims to
assess whether different cavity detection algorithms identify similar binding sites and
to validate whether the catalytic site identified through the reactivity-based method
corresponds to known catalytic residues. Additionally, HocusPockets seeks to identify
and prioritize relevant protein targets for rational drug design, using validation datasets
of previously categorized druggable and non-druggable proteins. The tool is
implemented in Python 3.10.12 on Ubuntu 22.04.4 LTS and will be publicly available
as open-source software.