Misannotations of rRNA can now generate 90% false positive protein matches in metatranscriptomic studies

Abstract

In the course of analyzing 9 522 746 pyrosequencing reads from 23 stations in the Southwestern Pacific and equatorial Atlantic oceans, it came to our attention that misannotations of rRNA as proteins is now so widespread that false positive matching of rRNA pyrosequencing reads to the National Center for Biotechnology Information (NCBI) non-redundant protein database approaches 90%. One conserved portion of 23S rRNA was consistently misannotated often enough to prompt curators at Pfam to create a spurious protein family. Detailed examination of the annotation history of each seed sequence in the spurious Pfam protein family (PF10695, ‘Cw-hydrolase’) uncovered issues in the standard operating procedures and quality assurance programs of major sequencing centers, and other issues relating to the curation practices of those managing public databases such as GenBank and SwissProt. We offer recommendations for all these issues, and recommend as well that workers in the field of metatranscriptomics take extra care to avoid including false positive matches in their datasets.

Metatranscriptome analysis of the reef-building coral Orbicella faveolata indicates holobiont response to coral disease

“…RRT expression might play a role in introducing genomic variation (Lampson et al, 2005) that influences bacterial gene expression, which would provide an avenue for microbiome adaptation under changing environmental conditions. At the same time, misannotations of rRNAs as proteins seems to be a widespread phenomenon associated with metatranscriptomics (Tripp et al, 2011). In particular, RRTs seem to be commonly misannotated 23S rRNA genes (Tripp et al, 2011), which would explain their abundance in our data.…”

“…At the same time, misannotations of rRNAs as proteins seems to be a widespread phenomenon associated with metatranscriptomics (Tripp et al, 2011). In particular, RRTs seem to be commonly misannotated 23S rRNA genes (Tripp et al, 2011), which would explain their abundance in our data. Hence, the bacterial RRT signature in our data should be treated with caution, and it remains to be determined to what extend increased retroelement activity is a shared characteristic between bacteria and the eukaryotic compartment.…”

Species radiation by DNA replication that systematically exchanges nucleotides?

“…The human sequence is 94% identical to the fungal sequence, suggesting that the mRNA originates from mammalian contamination. In addition, it is probable that it does not code for a protein, and is a misannotated (human) rRNA fragment (Tripp et al, 2011). Independently of these considerations on contamination and misannotation, the sequence is annotated 'mRNA'.…”

AntiFam: a tool to help identify spurious ORFs in protein annotation

“…Evidence for homology to known protein on opposite strand60PF07633 (DUF1594)15.0Proteins may not be expressed. Evidence for homology to known protein on opposite strand50PF11370 (DUF3170)25.0Probable shadow ORF of Clp protease167PF11194 (DUF2825)25.0Probable CRISPR b repeat regions15918PF11664 (DUF3264)25.0Probable CRISPR repeat regions2113PF10695 (Cw-hydrolase)25.0Antisense to rRNA (9)2251,654PF10919 (DUF2699)26.0Shadow ORF of PF00665 (integrase core domain 1)2511PF07641 (DUF1596)26.0Dubious genome annotation. Family comprises only three sequences from Rhodopirellula baltica , two overlapping30

The final two columns show the number of matches of each AntiFam entry to UniProtKB and to a metagenomic data set.

a The metagenomic set of sequences is the same as that used by Pfam (14).

…”