Machine learning-guided anti-photoaging peptides from Chinese giant salamander skin: Efficient preparation and...

Collagen peptides are ubiquitously applied in food systems for their versatile bioactivities but face constraints from labor-intensive enzymatic screening and zoonotic risks from terrestrial sources. This study developed machine learning (ML) models to optimize enzymatic hydrolysis of Chinese giant salamander skin (GSS) for anti-photoaging peptides. An artificial neural network (ANN) outperformed random forest models in predicting elastase inhibition rate (EIR) (ANN R² = 0.96 vs. RF R² = 0.82). Genetic algorithm-optimized hydrolysates achieved 59.33 % EIR at 100 mg/mL. In UVB-irradiated HaCaT cells, 2 mg/mL hydrolysate elevated type I collagen and elastin by 279 % and 274 %, respectively, while 40 μg/mL enhanced HSF cell expression by 41 % (collagen) and 69 % (elastin). Mechanistic analysis identified the key tetrapeptide PFGI as a regulator of MMP2 (matrix metalloproteinase 2) inhibition and EGFR (epidermal growth factor receptor) activation via estrogen signaling pathway. This ML-driven approach efficiently yields GSS-derived collagen peptides as effective ingredients for skin-health functional foods while addressing safety concerns.