Investigation of MiSeq reproducibility on biomarker identification

Abstract

MiSeq-derived artificial sequences appeared to be of good quality, thus bioinformatics tools failed to remove MiSeq artefacts. Even after removing singleton sequences or operational taxonomic units (OTUs), it is not clear how many sequence artefacts remained. Here, 16S rRNA genes were amplified from soil, human feces, pig feces, and groundwater. These were sequenced with five separate runs of MiSeq. Subsequently, each run of MiSeq was compared through alpha and beta-diversity analyses. We found more than half the OTUs were not in consensus through the multiple MiSeq runs, resulting in varying group-specific biomarker OTUs in each MiSeq run. Thus, differential abundance test should be interpreted with caution, and we suggest that results also should be verified further with other quantification methods such as qPCR.