Large-scale estimation of bacterial and archaeal DNA prevalence in metagenomes reveals biome-specific patterns.

Metagenomes often contain many reads derived from eukaryotes, but there is usually no reliable method for estimating their prevalence. This forces many analysis techniques to make the often-faulty assumption that all reads are prokaryotic. Here, we present SingleM prokaryotic_fraction (SPF), an algorithm that scalably and robustly estimates the number of bacterial and archaeal reads in a metagenome. It also estimates the average genome size of bacteria/archaea in a sample. SPF does not use eukaryotic reference genome data and can be applied to any modern Illumina metagenome. Based on SPF, we propose the domain-adjusted mapping rate (DAMR) as an improved metric to assess prokaryotic genome recovery from metagenomes. Applying SPF to 136,284 publicly available metagenomes, we report substantial variation in prokaryotic fractions and biome-specific patterns of prokaryotic abundance, providing insights into how microorganisms and eukaryotes are distributed across Earth. Finally, we show that substantial amounts of human host DNA sequence data have been deposited in public metagenome repositories, possibly counter to ethical directives that mandate screening of these reads prior to release. As the adoption of metagenomic sequencing continues to grow, we foresee SPF being a valuable tool for the appraisal of genome recovery efforts and for investigating global patterns of microorganism distribution.IMPORTANCEMetagenomics data sets capture DNA from all organisms in a sample, enabling the analysis of communities without relying on culture-based techniques. However, many samples include uncharacterized eukaryotic organisms and viral elements, meaning the proportion of bacterial and archaeal DNA is often unknown. This study presents SingleM prokaryotic_fraction (SPF), a robust and scalable method for estimating the prevalence of bacterial and archaeal DNA in metagenomes. Crucially, SPF is calculated independent of eukaryotic and viral reference genomes, which are often incomplete or unavailable. Applying SPF to over 136,000 public metagenomes uncovered substantial variability between microbial communities living in different environments. SPF also identified previously overlooked human genetic data contamination in public data sets, raising important ethical and privacy considerations. Building on SPF, we propose the domain-adjusted mapping rate (DAMR) metric, a new metric that improves genome recovery assessment by accounting for non-prokaryotic reads.