Quantifying microbial DNA in metagenomes improves microbial trait estimation

Shotgun metagenomics is a powerful tool for studying the genomic traits of microbial community members, such as genome size, gene content, etc. While such traits can be used to better understand the ecology and evolution of microbial communities, the accuracy of their estimations can be critically influenced by both known and unknown factors. One factor that can bias trait estimations is the proportion of eukaryotic and viral DNA in a metagenome, as some bioinformatic tools assume that all DNA reads in a metagenome are bacterial or archaeal. Here, we add to a recent debate about the influence of eukaryotic DNA in the estimation of average genome size from a global soil sample dataset using a new bioinformatic tool. Contrary to what was assumed, our reanalysis of this dataset revealed that soil samples can contain a substantial proportion of non-microbial DNA, which severely inflated the original estimates of average genome size. Correcting for this bias significantly improves the statistical support for the negative relationship between average bacterial genome size and soil pH. These results highlight that metagenomes can contain large quantities of non-microbial DNA and that new methods that correct for this can improve microbial trait estimation.