Annotation and evidence based functional curation of RefSeq prokaryotic genomes
Year: 2016
Type: other
Abstract: The National Center for Biotechnology Information (NCBI) has developed a robust prokaryotic genome annotation pipeline which is offered as a service to GenBank submitters and is used to annotate RefSeq prokaryotic genomes. NCBI’s annotation pipeline integrates annotation quality standards which were developed in part through a series of microbial assembly and annotation workshops held by NCBI. These workshops defined annotation quality criteria for complete genomes, standards for reporting support evidence, and protein naming protocols which are used by UniProt, GenBank, and RefSeq for eukaryotic and prokaryotic genomes. In 2015 the RefSeq project implemented a number of changes impacting the prokaryotic genomes data set. These include: a) development of a new framework for evidence based protein naming, name curation, and evidence tracking; b) increased collaboration to improve protein names; c) completion of the transition to a new data model for managing prokaryotic protein sequences; d) expanded assembly and annotation quality testing; e) increased capacity in NCBI’s prokaryotic genome annotation pipeline; and f) comprehensive reannotation of all RefSeq prokaryotic genomes. These developments expand on the established annotation QA, evidence, and name standards. As a result, RefSeq provides significant annotation consistency which facilitates comparative genomics. Our more rigorous QA criteria resulted in the suppression of over two thousand RefSeq prokaryotic genomes that do not meet these criteria, thus ensuring continued high quality of the RefSeq prokaryotic genomes dataset. Based on our new infrastructure to support protein name curation, we have established evidence-based and tracked names for approximately 25% of the RefSeq proteins thus far. Curation of protein names and name evidence uses a multi-faceted approach that leverages Hidden Markov Models (HMMs), domain architectures, available curated name data from Swiss-Prot and the Enzyme Commission, collaboration, and curation by NCBI staff. We use HMMs from several sources including TIGRfams and Pfams, and are creating new HMMs (NCBIfams) when further refinement is needed. The RefSeq group collaborates with NCBI’s Conserved Domains Database (CDD) group to provide functional names based on reviewed domain architectures. We collaborate with individual scientists and expert databases to provide the best names for some classes of proteins. NCBI staff curate name information both at the level of the support evidence (HMMs), and at the level of individual RefSeq protein names. Database tables support tracking the name update history, the biocurator or collaborator source of the update, and all support evidences for the name at the time of the update. We plan to start reporting functional evidence information on RefSeq protein records in 2016. The presentation will summarize the current RefSeq prokaryotic genomes process flows for genome annotation and curation/collaboration, our current assembly and annotation quality criteria, examples of annotation improvements resulting from collaboration and NCBI staff curation, and a proposal for reporting sets of functional evidence in the context of NCBI sequence displays.
Institution: National Center for Biotechnology Information
Language: English
Subfield: Molecular Biology
Domain: Life Sciences
Open Access status: gold
OpenAlex ID: W2731293501