A deep learning framework for oligopeptide candidate discovery across metabolism-related contexts.

Abstract

Artificial intelligence (AI)-based methods are increasingly critical in therapeutic peptide discovery but have been concentrated in certain applications, such as the development of antimicrobial peptides. Therapeutic peptide discovery in contexts such as metabolism, endocrinology, and tissue regeneration remains challenging due to a paucity of available indication-specific training data. Here, we propose a deep learning-based pipeline, Deepeptide, capable of identifying candidate oligopeptides for various metabolism-related contexts. Leveraging the intrinsic relationships between disease indication, biological processes, and molecular function, Deepeptide identifies oligopeptides associated with disease-related processes as lead candidates. Deepeptide was applied in five representative disease-related contexts: angiogenesis, lipid metabolism, osteogenesis, glucose metabolism, and anti-angiogenesis. Overall, 62% of the identified oligopeptide candidates demonstrated significant bioactivity, with most of them showing comparable potency to the benchmark therapeutic agents. These findings highlight the potential utility and generalizability of Deepeptide for oligopeptide lead discovery across metabolism-related contexts.