In addition, 33% of the studied sequences were erroneously annotated though missing fundamental catalytic residues.
This is due to the fact that first scoring hits (UniProtKB accession numbers are: A1CBG5, A1DDX4, Q2U688, Q6M9J3, A5DUJ4, Q4WVI2) were erroneously annotated as translation factors with GO terms belonging to the IEA category (as of March 2008).
A sequence similar protein from Lactococcus lactis, gi12723655, aligning only to the N-terminal module is erroneously annotated as having both aspartokinase and homoserine dehydrogenase activities.
The validation process was also hindered by genomic sequence repetitions, but with manual analysis, it was possible to identify intron sequences that had been erroneously annotated (Supplementary Table S2).
Of the 13 Arabidopsis genes that were erroneously annotated (Additional file 1), 12 show similarities to Picea glauca genes which are expressed late in embryogenesis and therefore named LEA despite the fact that they have different structural domains.
In addition, two clusters identified as an Aspartate-tRNA ligase (aspS, MSC_0432) and a CTP synthetase (pyrG, MSC_0902) are in fact partial sequences that were erroneously annotated and should also be considered as pseudogenes.
A plausible working hypothesis is that most of these sequences are in fact non-coding RNAs (ncRNAs) similar to mRNA-like ncRNAs [ 5, 6] that were erroneously annotated as protein coding genes.
Since several processed pseudogenes inserted into the genome very recently could be erroneously annotated as functional genes in Ensembl [ 48], we identified the single-exon genes with protein sequence identity ≥98% to multiple-exon genes as processed pseudogenes.
Upon inspection of the images falsely classified in the test sets, either as false positives (i.e. classified as gene expression in an anatomical component while the manual curation states that no expression is present) or false negatives, we found that several images were erroneously annotated.
Starmer et al. explained genome annotation errors with a bias in the start codon prediction towards the usage of GUG instead of AUG [ 27], whereas a previous study performed by Ma et al. [ 16] found in E. coli K12 a significant group of genes which started with GUG or UUG and which do not have an SD sequence and hence were erroneously annotated as putative or hypothetical proteins.
