Explainable AI for end-to-end pathogen target discovery and molecular design

Abstract

Drug discovery is limited by target identification, a major bottleneck in antimicrobial development and in combating emerging fungicide resistance. We present APEX (Attention-based Protein EXplainer), an explainable AI framework for cross-species, proteome-scale target discovery and pocket-guided molecular design. APEX combines ESM-2 embeddings, GAT, and a MLP to train pathogen-specific essentiality predictors (APEX-Tar) alongside a universal druggability model (APEX-Drug). Attention maps and GNNExplainer-derived subgraphs highlight residues driving predictions, enabling direct conditioning of structure-based diffusion models for inhibitor generation. APEX-Tar identifies key residues in known fungal targets and nominates new candidates, including adenylosuccinate lyase (ADSL) in fungi and the bacterial adhesin YadV. APEX-Drug recapitulates established fungicide binding sites, guides the design of ADSL inhibitors exploiting a fungal-specific active-site residue, and uncovers in YadV a previously undescribed pocket distinct from known pilicide sites. Together, APEX provides a kingdom-agnostic, explainable pipeline for target prioritization and guided molecular design, accelerating the search for next-generation antimicrobials.