Expert Curation in UniProtKB

The UniProt KnowledgeBase (UniProtKB) provides a single, centralized, authoritative resource for protein sequences and functional information. The majority of its records is based on automatic translation of coding sequences (CDS) provided by submitters at the time of initial deposition to the nucleotide sequence databases (INSDC).  At this point, all the information included in the original submission, even the erroneous one, is transferred to the record. Selected entries are then manually annotated. Besides integrating, interpreting and standardizing data from literature and numerous resources, curators are also checking, and often correcting, gene model predictions. For plants, this task is limited to Arabidopsis thaliana and Oryza sativa subsp. japonica.

UniProtKB has responded to the increasing submission of multiple versions of sequences from similar organisms and strains by developing the concept of complete and reference proteomes. Frequently, only the raw sequence of a complete genome is deposited to the nucleotide sequence databases and the gene model predictions and annotations are kept in separate, specialized model organism databases (MODs). UniProtKB had to implement pipelines for import of protein sequences from INSDC, Ensembl, EnsemblGenomes  and RefSeq as sources for the underlying complete genomes.

Data from high-throughput proteomics experiments constitute a rich potential source of annotations for UniProtKB. Experimental peptides that uniquely match the product of a single gene are used to generate annotations describing post-translational modifications and protein processing events, and to confirm protein existence. The knowledgebase is cross-referenced to numerous proteomics resource.