Identification of potent antimicrobial peptides via a machine-learning pipeline that mines the entire space of peptide sequences

Abstract

Systematically identifying functional peptides is difficult owing to the vast combinatorial space of peptide sequences. Here we report a machine-learning pipeline that mines the hundreds of billions of sequences in the entire virtual library of peptides made of 6-9 amino acids to identify potent antimicrobial peptides. The pipeline consists of trainable machine-learning modules (for performing empirical selection, classification, ranking and regression tasks) assembled sequentially following a coarse-to-fine design principle to gradually narrow down the search space. The leading three antimicrobial hexapeptides identified by the pipeline showed strong activities against a wide range of clinical isolates of multidrug-resistant pathogens. In mice with bacterial pneumonia, aerosolized formulations of the identified peptides showed therapeutic efficacy comparable to penicillin, negligible toxicity and a low propensity to induce drug resistance. The machine-learning pipeline may accelerate the discovery of new functional peptides.