Mining Bovine Milk Proteins for DPP-4 Inhibitory Peptides Using Machine Learning and Virtual Proteolysis

Area of Science:

• Biochemistry
• Bioinformatics
• Computational Biology

Background:

• Dipeptidyl peptidase-IV (DPP-4) enzyme inhibitors are a key therapeutic class for managing type 2 diabetes.
• Bioactive peptides from bovine milk proteins show promise as natural DPP-4 inhibitors.

Purpose of the Study:

• To develop and validate a combined machine learning and virtual proteolysis strategy for identifying DPP-4 inhibitory peptides.
• To discover and characterize novel DPP-4 inhibitory peptides from bovine milk proteins.

Main Methods:

• Training and evaluation of five machine learning models (GBDT, XGBoost, LightGBM, CatBoost, RF) for DPP-4 inhibitory activity prediction.
• Virtual proteolysis of milk proteins followed by in silico screening using the best-performing model (LightGBM).
• In vitro validation, molecular docking, and molecular dynamics simulations of identified peptides.

Main Results:

• LightGBM achieved a high AUC of 0.92 ± 0.01 in predicting DPP-4 inhibitory potential.
• The peptides GPVRGPF and HPHPHL demonstrated significant DPP-4 inhibitory activity.
• GPVRGPF was traced to β-casein (chymotrypsin hydrolysis), and HPHPHL to κ-casein (stem bromelain/papain hydrolysis).

Conclusions:

• The integration of machine learning and virtual proteolysis offers an efficient pipeline for discovering bioactive peptides.
• This strategy aids in predicting optimal enzymatic hydrolysis parameters for peptide generation.
• Identified peptides show potential for development into novel DPP-4 inhibiting diabetes medications.