The Limits and Avoidance of Biases in Metagenomic Analyses of Human Fecal Microbiota

An increasing body of evidence highlights the role of fecal microbiota in various human diseases. However, more than two-thirds of fecal bacteria cannot be cultivated by routine laboratory techniques. Thus, physicians and scientists use DNA sequencing and statistical tools to identify associations between bacterial subgroup abundances and disease. However, discrepancies between studies weaken these results. In the present study, we focus on biases that might account for these discrepancies. First, three different DNA extraction methods (G'NOME, QIAGEN, and PROMEGA) were compared with regard to their efficiency, i.e., the quality and quantity of DNA recovered from feces of 10 healthy volunteers. Then, the impact of the DNA extraction method on the bacteria identification and quantification was evaluated using our published cohort of sample subjected to both 16S rRNA sequencing and whole metagenome sequencing (WMS). WMS taxonomical assignation employed the universal marker genes profiler mOTU-v2, which is considered the gold standard. The three standard pipelines for 16S RNA analysis (MALT and MEGAN6, QIIME1, and DADA2) were applied for comparison. Taken together, our results indicate that the G'NOME-based method was optimal in terms of quantity and quality of DNA extracts. 16S rRNA sequence-based identification of abundant bacteria genera showed acceptable congruence with WMS sequencing, with the DADA2 pipeline yielding the highest congruent levels. However, for low abundance genera (<0.5% of the total abundance) two pipelines and/or validation by quantitative polymerase chain reaction (qPCR) or WMS are required. Hence, 16S rRNA sequencing for bacteria identification and quantification in clinical and translational studies should be limited to diagnostic purposes in well-characterized and abundant genera. Additional techniques are warranted for low abundant genera, such as WMS, qPCR, or the use of two bio-informatics pipelines.