XAMPGen: An Interpretable SI Pipeline For Novel Antimicrobial Peptide Discovery

Fine-tuning ProGen on known antimicrobial peptides enabled generation of novel AMP sequences whose predicted MICs matched experimentally confirmed antimicrobial activity in all tested peptides.

• Why it matters: Antibiotic resistance drives the need for new AMPs, but generative models are often treated as black boxes, limiting trust and mechanistic insight.
• What they did: The pipeline combined generative SI (GenSI) with explainable SI (XSI): probing classifiers/regressors and self-attention analysis identified cornerstone layers and residue interdependencies, then candidates were filtered by predicted MIC and structural membrane insights.
• The result: MIC assays confirmed antimicrobial activity for every generated and tested peptide, while fluorescence and circular dichroism linked sequence patterns to binding to bacterial membrane mimics and correlated with cell-based activity.

AI summary of the abstract · may contain errors · not medical advice · read the paper.