LIGMAP

This proposal describes LIGMAP, a research project developing advanced AI methods for characterising and predicting protein pockets and their natural ligand interactions across the human proteome. The work integrates modern pocket detection algorithms, evolutionary analysis, and machine learning to systematically assign endogenous metabolites or cofactors to protein binding sites. By doing so, the proposal seeks to illuminate the roles of uncharacterised proteins, trace metabolic and signalling pathways, explain disease-associated mutations, and support drug discovery and enzyme engineering. A central innovation is EvoPockets, a new approach that combines geometric, physicochemical, and evolutionary data to identify and analyse binding sites that are conserved and biologically significant. The project further leverages evolutionary conservation and protein family clustering to improve ligand prediction and functional annotation. LIGMAP is organised into five interconnected work packages, focusing on pocket identification, structural modelling, pocket characterisation, AI-based protein ligand prediction, and open data access. LIGMAP will provide a comprehensive, openly accessible framework linking protein structure to biological function and disease, providing valuable tools for researchers in academia, healthcare, and industry. We will report endogenous ligands for human, viral and Neglected Tropical Disease pathogen proteomes. Thus, LIGMAP will advance understanding of protein function and facilitate discoveries in biology and medicine.