Bacterial genomes, new protein data model and accession series (WP)

NCBI continues to expand the RefSeq bacterial genomes node to include ALL complete and draft genomes that meet minimum assembly and annotation quality criteria. This means that RefSeq will include more than one genome of the same strain which may be provided through strain population sampling or sequencing to monitor a disease outbreak. NCBI is in the process of re-annotating all bacterial genomes, with the exception of a small umber for which annotation is provided by, or in collaboration with, another group (such as E. coli str. K12 substr. MG1655).

Due to the expanded scope of the RefSeq bacterial node, we anticipated a very large increase in the number of identical (redundant) proteins; therefore, we have introduced a new data model for bacterial proteins whereby we are providing a true non-redundant protein dataset associated with a new accession prefix,'WP'. Details about the new data model with examples was announced between release cycles.

This release includes a new supplemental file providing mapping of WP accessions to tax_id and species name, for the subset of WP accessions that are annotated on genomes of different species. For example, see WP_000002243.1. The mapping file is available in the release-catalog directory.

We strongly encourage you to read the full announcement.

Supression of some bacterial genomes

Please note that some RefSeq bacterial genomes were recently suppressed. This includes unannotated genomes that had not been processed by NCBI's annotation pipeline yet and annotated genomes with identified annotation quality issues. This has resulted in a net decrease in RefSeq bacterial genomic accessions in this release. Many of the suppressed accessions will be reinstated when annotation is provided.

Changes in annotation of human and vertebrate transcript records

Recent changes to human and other vertebrate transcript records includes:

• removal of exon numbers
• expanded reporting of support evidence, in a structured comment with the header 'Evidence Data'
• (new) reporting gene and transcript attributes, in a structured comment with the header 'RefSeq Attributes'
• removal of mitochondrial localization information from the record DEFINITION line (moved to Attributes)

Please see the detailed description of these changes.

Policy change to allow a mixture of known and model accessions for eukaryotic genes

Previously, we did not allow a mixture of X* series accessions (genome annotation models) and N* series accessions (based on cDNA and curation) for a gene. We have changed this policy in order to provide increased annotation of splice variants. RefSeq models are calculated using cDNA, protein, and RNAseq data. There may be good support at the level of each exon pair; however, the long range exon combination represented in the model may not be fully supported and thus is less likely to be represented with a N* series accession. For example, see Gene ID: 100306968.