AMPpred-CNN: Prediction of Antimicrobial Peptide by Using 1D Convolution Neural Network and Composition/Transition/Distribution (CTD) Encoding

Antimicrobial resistance has become a global health crisis due to antibiotic misuse. Antimicrobial peptides (AMPs) show promise with broad pathogenic activity. However, experimental AMP discovery is expensive, and computational methods face challenges with data availability and resistance patterns. We introduce a method AmpPred-CNN based on 1D convolutional neural network and peptide encoding for their computational identification. Validated on a diverse dataset of 3268 AMPs and 166,791 non-AMPs, our model achieved impressive accuracy of 96%. Comparative evaluations confirmed its superiority. Notably, even with just the top 10 features, an accuracy of 94% is obtained. Additionally, AMP’s antimicrobial activities were found independent of sequence length. These findings underscore the significant improvement in AMP prediction accuracy through effective peptide encoding techniques and diverse training datasets, offering hope in combating antimicrobial resistance.