AI-driven discovery of multifunctional peptides: From sequence space to therapeutics.

AI-driven deep learning and protein language models have enabled data-driven, scalable identification of multifunctional peptides, improving representation of contextual determinants for MFP discovery.

• Why it matters: Conventional high-throughput screening and rule-based computation are labor-intensive, costly, low-yield, and ill-suited to the vast combinatorial peptide sequence space.
• What they did: This review systematically examines methodological innovations—peptide representation learning, multi-modal fusion, multi-label learning, and ProtLM-based embeddings—and summarizes applications in database mining, mechanism interpretation, and mutation effect prediction.
• The result: It highlights persistent challenges—class imbalance, incomplete annotation, limited cross-dataset generalizability, and insufficient integration of drug-likeness and developability criteria—while outlining promising directions for progress.

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