AI-assisted discovery of dual antioxidant and ACE-inhibitory peptides from Hericium erinaceus.

Hericium erinaceus harbored rich peptides, however their biofunctional properties remained unclear. In this study, Hericium erinaceus protein hydrolysates (HEph) were analyzed using a large language model (LLM)-assisted annotation workflow on the DeepSeek platform, combined with liquid chromatography-tandem mass spectrometry, antioxidant, angiotensin-converting enzyme (ACE)-inhibitory assays and cellular evaluations. Results showed that AI accurately annotated 7 functional peptides in HEph, and both these peptides and HEph exhibited anti-oxidant/ACE activities. All the annotated peptides were safe for human umbilical vein endothelial cells at tested doses, while certain ones modulated ROS, SOD, eNOS, and ET-1, VDYP and IHPTPESVFASHKHLFPPLP exhibited both dual anti-oxidant/ACE activities. Molecular docking indicated that peptide anti-oxidant/ACE activities were associated with the distribution of hydrophilic and aromatic residues. Overall, this study demonstrated the utility of LLMs assisted annotation for literature-grounded peptide curation and prioritization, uncovering atypical potential functional peptides, providing a strategy for the development of Hericium erinaceus peptides in functional foods.