Enhancing Antimicrobial Activity Predictors Based on Machine Learning Approaches

Abstract

Recently, the prediction tools of antimicrobial activity revealed a promising avenue for novel antimicrobial peptide (AMP) sequence determination and discovery. Machine learning (ML) approaches can be utilized to offer the prediction of AMP sequence with great success, which explores alternative strategies to combat antimicrobial resistance and develop effective treatments for infections. The main objective of this chapter is to study and evaluate the predictive ability of modern ML methods to accurately identify the activities of antimicrobial sequences previously described at the protein level through in vitro studies. To formally confirm whether the utilized ML approaches have a significant enhancement, the authors used a dataset with size 6623 instances for both AMP and non-AMP classes. The best performance was LGBM with an accuracy of 0.92%, MCC of 0.83, recall of 90%, Area Under the Curve (AUC) of 0.97%, precision of 0.91%, and F1-score of 0.92%.