Search

Microorganisms produce an immense variety of natural products through the expression of Biosynthetic Gene Clusters (BGCs): physically clustered genes that encode the enzymes of a specialized metabolic pathway. These natural products cover a wide range of chemical classes (e.g., aminoglycosides, lantibiotics, nonribosomal peptides, oligosaccharides, polyketides, terpenes) that are highly valuable for industrial and medical applications1. Metagenomics, as a culture-independent approach, has greatly enhanced our ability to survey the functional potential of microorganisms and is growing in popularity for the mining of BGCs. However, to effectively exploit metagenomic data to this end, it will be crucial to more efficiently identify these genomic elements in highly complex and ever-increasing volumes of data2. Here, we address this challenge by developing the ultrafast Biosynthetic Gene cluster MEtagenomic eXploration toolbox (BiG-MEx). BiG-MEx rapidly identifies a broad range of BGC protein domains, assess their diversity and novelty, and predicts the abundance profile of natural product BGC classes in metagenomic data. We show the advantages of BiG-MEx compared to standard BGC-mining approaches, and use it to explore the BGC domain and class composition of samples in the TARA Oceans3 and Human Microbiome Project datasets4. In these analyses, we demonstrate BiG-MEx’s applicability to study the distribution, diversity, and ecological roles of BGCs in metagenomic data, and guide the exploration of natural products with clinical applications.Competing Interest StatementThe authors have declared no competing interest.

Benchmarking the accuracy of acn.sh against PICRUSt, CopyRighter, and PAPRICA: figure illustration. Plots of the Pearsonâ s correlation coefficient (upper panel) and the absolute percentage error (APE) value distributions (lower panel) of the ACN computed by acn.sh, PICRUSt, CopyRighter, and PAPRICA, with respect to the reference ACN. As mentioned above, we compared these tools using simulated metagenomes of different read length of the General, Infant Gut, and Marine datasets. For the sake of clarity, 100 outlier APE values (4.2% of the total data) were not included in the plot. (PDF 62 kb)

Benchmarking the accuracy of ags.sh against MicrobeCensus: figure illustration. Plots of the Pearsonâ s correlation coefficients (upper panel) and absolute percentage error (APE) value distributions (lower panel) of the AGS computed by ags.sh and MicrobeCensus, with respect to the reference AGS. The comparisons were performed using the simulated metagenomes of different read length of the General, Infant Gut, and Marine datasets. For the sake of clarity, 70 outlier APE values (2.3% of the total data) were not included in the plot. (PDF 60 kb)