Trying to sew with a needle from a haystack: cautionary tales of assembly and taxonomy assignment from a White Oak River sediment metagenome

We analyzed a short read metagenome dataset from estuarine sediment in the White Oak River, NC to examine microbial community composition and functional potential. The dataset was assembled through multiple methods into longer reads, which were analyzed for functional and taxonomic content. While different assemblies yielded somewhat similar functional calls showing anaerobic metabolic pathways, the taxonomic calls varied depending on the pipeline applied. Homology-based assignment for taxonomy was most consistent with direct ribosomal RNA amplicon analysis, yet wide variations were seen between amplicon and metagenome identifications. Deep branching organisms may result in misidentification of genes through pipelines. We questioned whether assembly was creating chimeras and if fragment recruitment could assist in assembly. We show that close relatives may assist in assembly refinement, but that many relatives are missing for this environment in the databases. In total, we have a cautionary tale of the potential for bias and misidentification in metagenomic annotation, depending on analysis method