Abstract
Genome annotation errors are a persistent problem that impede research in the biosciences. A manual curation effort is described that attempts to produce high-quality genome annotations for a set of haloarchaeal genomes (Halobacterium salinarum and Hbt. hubeiense, Haloferax volcanii and Hfx. mediterranei, Natronomonas pharaonis and Nmn. moolapensis, Haloquadratum walsbyi strains HBSQ001 and C23, Natrialba magadii, Haloarcula marismortui and Har. hispanica, and Halohasta litchfieldiae). Genomes are checked for missing genes, start codon misassignments, and disrupted genes. Assignments of a specific function are preferably based on experimentally characterized homologs (Gold Standard Proteins). To avoid overannotation, which is a major source of database errors, we restrict annotation to only general function assignments when support for a specific substrate assignment is insufficient. This strategy results in annotations that are resistant to the plethora of errors that compromise public databases. Annotation consistency is rigorously validated for ortholog pairs from the genomes surveyed. The annotation is regularly crosschecked against the UniProt database to further improve annotations and increase the level of standardization. Enhanced genome annotations are submitted to public databases (EMBL/GenBank, UniProt), to the benefit of the scientific community. The enhanced annotations are also publically available via HaloLex.
Highlights
Protein function assignments in public databases suffer from severe errors
We describe an effort for a high-quality annotation of a set of haloarchaeal genomes
The annotation of Hbt. salinarum strain NRC-1 [45], the classical genome of halophilic archaea, is covered by our approach as most its genes are represented in strain R1 with an identical protein sequence
Summary
Protein function assignments in public databases suffer from severe errors. Genomes are commonly subjected to automatic annotation procedures by computational annotation robots. As these procedures build on the information provided in public databases, errors in the database may be “propagated, leading to a potential transitive catastrophe” [3]. Error propagation could be substantially reduced if annotations are copied only from those proteins which themselves have been functionally characterized. Such proteins are referred to as “Gold Standard Proteins” [4,5]. The SwissProt section of UniProt is a rich source for Gold Standard
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
