A Sub-OTU-Resolution Analysis of the Human Microbiome

Abstract

To identify the bacteria that reside within a sample, researchers commonly amplify and sequence the 16S rRNA gene (16S). 16S is a gene unique to the prokaryotic genome that can be used to make taxonomic assignments down to genus/species level. In 16S studies, sequences that are over 97% similar are typically clustered into Operational Taxonomic Units (OTUs) and are assigned a taxon based upon their OTU. However, the use of OTUs reduces taxonomic resolution. To avoid this reduced resolution, a new approach, Cluster Free Filtering (CFF), was developed at Princeton. CFF is a denoising algorithm that estimates the error rate associated with the sequence data, and removes sequences whose abundance falls within the estimated error rate, leaving a set of sequences that actually occur within samples, with sub-OTU resolution. CFF was applied to a comparative study of the urinary microbiomes of healthy individuals and patients diagnosed with Overactive Bladder Syndrome (OAB). The increased resolution facilitated the detection of significant differences in the bacterial communities of the two test cohorts. CFF was also applied to raw 16S data obtained by the Human Microbiome Project (HMP). The analysis yielded an updated “Most Wanted” Taxa list, namely, a list of bacterial species and genera highly associated with the human body but have not been thoroughly studied. In addition to yielding the “Most Wanted” Taxa list, it was found that the distribution of sequence abundances in a given sample could typically be well fit by a stretched exponential distribution. Moreover, it was observed that microbiomes along the oral tract and the digestive tract share a similar stretching exponent.