DPP-IV inhibitory peptides from highland barley via machine learning and multi-scale validation

Highland barley has shown potential in regulating blood glucose and may serve as a natural source of dipeptidyl peptidase-IV (DPP-IV) inhibitors. In this study, machine learning (Gradient Boosting Decision Trees) and virtual screening were employed to identify DPP-IV inhibitory peptides from highland barley protein. Three candidate peptides (FPQPQ, FPRPF, and YGGWN) had IC₅₀ of 675.47, 766.80, and 281.76 μM, respectively. FPQPQ and FPRPF are competitive DPP-IV inhibitors, whereas YGGWN is a non-competitive inhibitor. Their inhibitory mechanisms were investigated through molecular docking, and molecular dynamics simulations. Network pharmacology was applied to reveal their multitarget and multipathway antihyperglycaemic activities in vivo. This integrated approach enabled efficient and precise screening of bioactive peptides and provided mechanistic insights into their inhibitory effects. The findings demonstrate the potential of these peptides in glucose regulation and support the development of functional foods based on highland barley protein.