Aim: Hypertension is associated with development of cardiovascular disease and has become a significant health problem worldwide. Naturally-derived antihypertensive peptides have emerged as promising alternatives to synthetic drugs.
Materials & methods: This study introduces predictor of antihypertensive activity of peptides constructed using random forest classifier as a function of various combinations of amino acid, dipeptide and pseudoamino acid composition descriptors.
Results: Classification models were assessed via independent test set that demonstrated accuracy of 84.73%. Feature importance analysis revealed the preference of proline and hydrophobic amino acids at the C-terminal as well as the preference of short peptides for robust activity.
Conclusion: Model presented herein serves as a useful tool for predicting and analysis of antihypertensive activity of peptides.
Keywords: angiotensin-converting enzyme inhibitory peptides; antihypertensive peptides; hypertension; machine learning; random forest.