Accurate estimation of microbial diversity with Distanced

Abstract

Microbes are the most diverse organisms on Earth. Sequencing their DNA suggests thousands of different microbes could be present in a single sample. Errors in sequencing, however, make it challenging to estimate exactly how diverse microbes are. Here we developed a tool that estimates diversity accurately, even in the presence of sequencing errors. We first evaluated two existing tools, DADA2 and Deblur, which work by correcting sequencing errors. We found that these tools estimated within-sample (alpha) diversity poorly. In fact, we obtained better estimates if did not use the tools at all (left errors uncorrected). These tools performed poorly because they changed the relative abundance of different sequences; this is a side effect of correcting errors and discarding up to 90% of sequence reads in the process. Previous evaluations ignored sequence abundance when calculating diversity, overlooking this problem. Our tool, Distanced, differs from existing tools because it does not correct sequencing errors. Instead, it corrects sequence distances, which are used to calculate diversity. It does this correction with Phred quality scores and Bayes theorem. No sequence reads are discarded in the process. In our evaluation, Distanced accurately estimated diversity of bacterial DNA, fungal DNA, and even antibody mRNA. Given its accuracy, Distanced will help investigators answer important questions about microbial diversity. For example, it could answer how important is diversity for the planets ecosystems and human health.