Predicting the secondary metabolic potential of microbiomes from marker genes using PSMPA

Background: The efficient discovery of novel antibiotics is of great significance for us to fight against drug-resistant pathogens. Previously, a great deal of time and effort has been spent on screening and isolating novel antibiotic-producing bacteria from complex environmental samples, and the secondary metabolic potential of microbiomes could only be investigated after their genome sequences were available. Results: Here, we present PSMPA, a web server and a standalone tool, for predicting the numbers of each class of bacterial secondary metabolite biosynthetic gene clusters (BGCs) in environmental samples using 16S rRNA gene amplicons, which could prioritize samples and bacterial strains with high potential to produce novel antibiotics at an early stage. The pipeline integrated PICRUSt2 and BLASTn, and relied on a comprehensive bacterial BGC atlas which contains 1,295,905 BGCs from 216,408 bacterial genomes. PSMPA showed good performance with the accuracy largher than 80% when applied to predict the BGC profiles in 5,000 randomly selected bacterial genomes. Then, PSMPA was applied to depict the distribution of BGCs in microbiomes from human gut, sea water, deep-sea sediments, and soil samples from several independent datasets, which uncovered plenty of novel strains that are rich in BGCs. Conclusions: We presented a comprehensive bacterial BGC atlas and demonstrated that PSMPA is a usefull tool for predicting the secondary metabolic potential of microbiomes from marker genes. PSMPA would facilitate the efficient discovery of novel microbial secondary metabolites and enrich the resource for amplicon sequencing-based functional analysis. The PSMPA is available at https://www.psmpa.net.