Predicting Short Antimicrobial Peptides Utilizing Multi-tiered Stacked Ensemble Machine Learning Technique

Antimicrobial peptides (AMPs), a crucial component of innate immunity, have gained attention as potential candidates for novel antibiotics due to their broad-spectrum antimicrobial activity. Particularly, short AMPs offer significant advantages due to their swift response against pathogens, effectively combating a diverse array of bacteria, viruses, fungi, and even cancer cells. However, the traditional screening of such AMPs from large peptide libraries is costly and time-consuming. To address this challenge, computational methods based on machine learning (ML) have been increasingly utilized for predicting AMPs from genomic data. This study proposes a multi-tiered stacked ensemble ML technique for predicting short antimicrobial pep-tides. The research aims to leverage diverse ML algorithms and sequence encoding features to enhance prediction accuracy. Specifically, two types of features, Amino Acid Composition (AAC) and Dipeptide Composition (DPC), are extracted from a dataset of peptide sequences ranging from 5 to 30 amino acids. The dataset, containing positive and negative AMPs, is preprocessed, then trained and evaluated using various ML classifiers, including classical models, ensemble models (bagging and boosting), and deep neural networks (DNNs), with AAC and DPC features. Furthermore, a stacking ensemble model is proposed, wherein diverse base learners, such as LR, DT, SGD, SVM, and KNN, are combined with an MLP meta-learner. This stacked ensemble model demonstrates superior performance compared to individual ML models, achieving accuracy scores of 80.10% and 76.22% using AAC and DPC features, respectively. The study highlights the potential of stacked ensemble ML techniques for improving the prediction of short antimicrobial peptides. Future research directions include exploring alternative sequence encoding methods and expanding the dataset to further enhance prediction accuracy and robustness in identifying novel antimicrobial peptides.