Mining underutilized whole-genome sequencing projects to improve 16S rRNA databases

Current approaches to interpreting 16S rDNA amplicon data are hampered by several factors. Among these are database inaccuracy or incompleteness, sequencing error, and biased DNA/RNA extraction. Existing 16S rRNA databases source the majority of sequences from deposited amplicon sequences, draft genomes, and complete genomes. Most of the draft genomes available are assembled from short reads. However, repeated ribosomal regions are notoriously difficult to assemble well from short reads, and as a consequence the short-read-assembled 16S rDNA region may be an amalgamation of different loci within the genome. This complicates high-resolution community analysis, as a draft genomes 16S rDNA sequence may be a chimera of multiple loci; in such cases, the draft-derived sequences in a database may not represent a 16S rRNA sequence as it occurs in biology. We present Focus16, a pipeline for improving 16S rRNA databases by mining NCBIs Sequence Read Archive for whole-genome sequencing runs that could be reassembled to yield additional 16S rRNA sequences. Using riboSeed (a genome assembly tool for correcting rDNA misassembly), Focus16 provides a way to augment 16S rRNA databases with high-quality re-assembled sequences. In this study, we augmented the widely-used SILVA 16S rRNA database with the novel sequences disclosed by Focus16 and re-processed amplicon sequences from several benchmarking datasets with DADA2. Using this augmented SILVA database increased the number of amplicon sequence variants that could be assigned taxonomic annotations. Further, fine-scale classification was improved by revealing ambiguities. We observed, for example, that amplicon sequence variants (ASVs) may be assigned to a specific genus where Focus16-correction would indicate that the ASV is represented in two or more genera. Thus, we demonstrate that improvements can be made to taxonomic classification by incorporating these carefully re-assembled 16S rRNA sequences, and we invite the community to expand our work to augment existing 16S rRNA reference databases such as SILVA, GreenGenes, and RDP.