Abstract
The human gut microbiota is well established as an important factor in health and disease. Fecal sample microbiota are often analyzed as a proxy for gut microbiota, and characterized with respect to their composition profiles. Modern approaches employ whole genome shotgun next-generation sequencing as the basis for these analyses. Sequencing depth as well as choice of next-generation sequencing data analysis method constitute two main interacting methodological factors for such an approach. In this study, we used 200 million sequence read pairs from one fecal sample for comparing different taxonomy classification methods, using default and custom-made reference databases, at different sequencing depths. A mock community data set with known composition was used for validating the classification methods. Results suggest that sequencing beyond 60 million read pairs does not seem to improve classification. The phylogeny prediction pattern, when using the default databases and the consensus database, appeared to be similar for all three methods. Moreover, these methods predicted rather different species. We conclude that the choice of sequencing depth and classification method has important implications for taxonomy composition prediction. A multi-method-consensus approach for robust gut microbiota NGS analysis is recommended.