Identification of Antimicrobial Peptides from Macroalgae with Machine Learning

Abstract

Antimicrobial peptides (AMPs) are essential components of innate host defense showing a broad spectrum of activity against bacteria, viruses, fungi, and multi-resistant pathogens. Despite their diverse nature, with high sequence similarities in distantly related mammals, invertebrate and plant species, their presence and functional roles in marine macroalgae remain largely unexplored. In recent years, computational tools have successfully predicted and identified encoded AMPs sourced from ubiquitous dual-functioning proteins, including histones and ribosomes, in various aquatic species. In this paper, a computational design is presented that uses machine learning classifiers, artificial neural networks and random forests, to identify putative AMPs in macroalgae. 42,213 protein sequences from five macroalgae were processed by the classifiers which identified 24 putative AMPs. While initial testing with AMP databases positively identifies these sequences as AMPs, an absolute determination cannot be made without in vitro extraction and purification techniques. If confirmed, these AMPs will be the first-ever identified in macroalgae.