Tutorial: guidelines for the use of machine learning methods to mine genomes and proteomes for antibiotic discovery

Genomes and proteomes constitute a rich reservoir of molecular diversity. However, they have remained underexplored because of a lack of appropriate tools. In recent years, computational approaches have been developed to mine this unexplored biological information, or dark matter, accelerating the discovery of new antibiotic molecules. Such efforts have yielded a wide range of new molecules. These include peptides released via predicted proteolytic cleavage of larger proteins, termed 'encrypted peptides', which have been found to be widespread in nature. Molecules encoded by and translated from small open reading frames within genomic sequences have also been uncovered, further expanding the landscape of bioactive compounds. Here, we discuss computational approaches, including machine learning and artificial intelligence (AI) tools, which have been used to date to identify antimicrobial compounds, with a special emphasis on peptides. We also propose potential avenues for future exploration in this rapidly evolving field. Moreover, we provide an overview of the experimental methods commonly used to validate these computational predictions. We anticipate that efforts combining cutting-edge AI and experimental approaches for biological sequence mining will reveal new insights into host immunity and continue to accelerate discoveries in the fields of antibiotics and infectious diseases.