Toward standards in clinical microbiome studies: comparison of three DNA extraction methods and two bioinformatic pipelines

When studying the microbiome using next generation sequencing, DNA extraction method, sequencing procedures and bioinformatic processing are crucial to obtain reliable data. Method choice has been demonstrated to strongly affect the final biological interpretation. We assessed the performance of three DNA extraction methods and two bioinformatic pipelines for bacterial microbiota profiling through 16S rRNA gene amplicon sequencing, using positive and negative controls for DNA extraction and sequencing, and eight different types of high- or low-biomass samples. Performance was evaluated based on quality control passing, DNA yield, richness, diversity and compositional profiles. All DNA extraction methods retrieved the theoretical relative bacterial abundance with maximum three-fold change, although differences were seen between methods, and library preparation and sequencing induced little variation. Bioinformatic pipelines showed different results for estimating richness, but diversity and compositional profiles were comparable. DNA extraction methods were successful for feces and oral swabs and variation induced by DNA extraction methods was lower than inter-subject (biological) variation. For low-biomass samples, a mixture of genera present in negative controls and sample-specific genera, possibly representing biological signal, were observed. We conclude that the tested bioinformatic pipelines perform equally with pipeline-specific advantages and disadvantages. Two out of three extraction methods performed equally well, while one method was less accurate regarding retrieval of compositional profiles. Lastly, we demonstrate the importance of including negative controls when analyzing low bacterial biomass samples.IMPORTANCEMethod choice throughout the workflow of a microbiome study, from sample collection to DNA extraction and sequencing procedures, can greatly affect results. This study evaluated three different DNA extraction methods and two bioinformatic pipelines by including positive and negative controls, and various biological specimens. By identifying an optimal combination of DNA extraction method and bioinformatic pipeline use, we hope to contribute to increased methodological consistency in microbiome studies. Our methods were not only applied to commonly studied samples for microbiota analysis, e.g. feces, but also for more rarely studied, low-biomass samples. Microbiota composition profiles of low-biomass samples (e.g. urine and tumor biopsies) were not always distinguishable from negative controls, or showed partial overlap, confirming the importance of including negative controls in microbiome studies, especially when low bacterial biomass is expected.