Abstract

International audience; Targeted metagenomics, also known as metagenetics, is a high-throughput sequencing application focusing on a nucleotide target in a microbiome to describe its taxonomic content. A wide range of bioinformatics pipelines are available to analyze sequencing outputs, and the choice of an appropriate tool is crucial and not trivial. No standard evaluation method exists for estimating the accuracy of a pipeline for targeted metagenomics analyses. This article proposes an evaluation protocol containing real and simulated targeted metage-nomics datasets, and adequate metrics allowing us to study the impact of different variables on the biological interpretation of results. This protocol was used to compare six different bioinformatics pipelines in the basic user context: Three common ones (mothur, QIIME and BMP) based on a clustering-first approach and three emerging ones (Kraken, CLARK and One Codex) using an assignment-first approach. This study surprisingly reveals that the effect of sequencing errors has a bigger impact on the results that choosing different amplified regions. Moreover, increasing sequencing throughput increases richness overestima-tion, even more so for microbiota of high complexity. Finally, the choice of the reference database has a bigger impact on richness estimation for clustering-first pipelines, and on correct taxa identification for assignment-first pipelines. Using emerging assignment-first pipelines is a valid approach for targeted metagenomics analyses, with a quality of results comparable to popular clustering-first pipelines, even with an error-prone sequencing technology like Ion Torrent. However, those pipelines are highly sensitive to the quality of databases and their annotations, which makes clustering-first pipelines still the only reliable approach for studying microbiomes that are not well described. Document type: Article

Abstract

Discovery of novel viruses in host samples is a multidisciplinary process which relies increasingly on next-generation sequencing (NGS) followed by computational analysis. A crucial step in this analysis is to separate host sequence reads from the sequence reads of the virus to be discovered. This becomes especially difficult if no reference genome of the host is available. Furthermore, if the total number of viral reads in a sample is low, de novo assembly of a virus which is a requirement for most existing pipelines is hard to realize. We present a new modular, computational pipeline for discovery of novel viruses in host samples. While existing pipelines rely on the availability of the hosts reference genome for filtering sequence reads, our new pipeline can also cope with cases for which no reference genome is available. As a further novelty of our method a decoy module is used to assess false classification rates in the discovery process. Additionally, viruses with a low read coverage can be identified and visually reviewed. We validate our pipeline on simulated data as well as two experimental samples with known virus content. For the experimental samples, we were able to reproduce the laboratory findings. Our newly developed pipeline is applicable for virus detection in a wide range of host species. The three modules we present can either be incorporated individually in other pipelines or be used as a stand-alone pipeline. We are the first to present a decoy approach within a virus detection pipeline that can be used to assess error rates so that the quality of the final result can be judged. We provide an implementation of our modules via Github. However, the principle of the modules can easily be re-implemented by other researchers.