Rapid and accurate taxonomic classification of cpn60 amplicon sequence variants

The universal target region of the gene encoding the 60 kDa chaperonin protein (cpn60, also known as groEL or hsp60) is a proven sequence barcode for bacteria and a useful target for marker gene amplicon-based studies of complex microbial communities. To date, identification of cpn60 sequence variants from microbiome studies has been accomplished by alignment of queries to a reference database. Naive Bayesian classifiers offer an alternative identification method that provides variable rank classification and shorter analysis times. We curated a set of cpn60 barcode sequences to train the RDP classifier and tested its performance on data from previous human microbiome studies. Results showed that sequences accounting for 79%, 86% and 92% of the observations (read counts) in saliva, vagina and infant stool microbiome data sets were classified to the species rank. We also trained the QIIME 2 q2-feature-classifier on cpn60 sequence data and demonstrated that it gives results consistent with the standalone RDP classifier. Successful implementation of a naive Bayesian classifier for cpn60 sequences will facilitate future microbiome studies and open opportunities to integrate cpn60 amplicon sequence identification into existing analysis pipelines.

Automated summary

The gene encoding the 60 kDa chaperonin protein (cpn60) is a reliable barcode for bacteria and a useful target for studying microbial communities. Current methods for identifying cpn60 sequence variants rely on alignment to a reference database, but a naive Bayesian classifier offers a faster alternative. In this study, we trained a classifier using curated cpn60 sequence data and successfully classified a high percentage of sequences in microbiome datasets.