A microbiome reality check: limitations of in silico-based metagenomic approaches to study complex bacterial communities

Abstract

In recent years, whole shotgun metagenomics (WSM) of complex microbial communities has become an established technology to perform compositional analyses of complex microbial communities, an approach that is heavily reliant on bioinformatic pipelines to process and interpret the generated raw sequencing data. However, the use of such in silico pipelines for the microbial taxonomic classification of short sequences may lead to significant errors in the compositional outputs deduced from such sequencing data. To investigate the ability of such in silico pipelines, we employed two commonly applied bioinformatic tools, i.e., MetaPhlAn2 and Kraken2 together with two metagenomic data sets originating from human and animal faecal samples. By using these bioinformatic programs that taxonomically classify WSM data based on marker genes, we observed a trend to depict a lower complexity of the microbial communities. Here, we assess the limitations of the most commonly employed bioinformatic pipelines, i.e., MetaPhlAn2 and Kraken2, and based on our findings, we propose that such analyses should ideally be combined with experimentally based microbiological validations.