Machine learning-driven discovery of bioactive peptides from duckweed (Lemnaceae) protein hydrolysates: Identification and experimental validation of 20 novel antihypertensive, antidiabetic, and/or antioxidant peptides

Abstract

Duckweed, a sustainable, protein-rich aquatic plant, has recently emerged as a promising source of bioactive peptides. However, their identification remains limited and challenging in such complex mixtures. Following duckweed hydrolysis with pepsin, chymotrypsin, trypsin and papain, and a centrifugation step producing two fractions: supernatant (DS) and pellet (DP), interesting half-maximal inhibitory concentration (IC₅₀) for dipeptidyl peptidase (DPP)-IV and angiotensin-converting enzyme (ACE) inhibition were obtained for DS fractions, especially with pepsin (IC₅₀ = 0.7 and 0.07 mg/mL, respectively). Using partial least squares-discriminant analysis (PLS-DA) combined with quantitative structure-activity relationship (QSAR) models, five new DPP-IV inhibitors (most active: API, IC₅₀ = 126.88 μM), eleven new ACE inhibitors (most active: FAR, IC₅₀ = 13.54 μM) and four new antioxidants (>200 μM) were identified. Two sequences were active across all three tested bioactivities, revealing promising multi-target peptides. These findings highlight the potential of duckweed-derived peptides to support health and metabolic balance.