the gene, and/or relevant publications. Manual curation and literature review by the RefSeq group can result in the representation of unique variants and isoforms that would not be predicted when based solely on computational analysis. For instance, literature review of the human tumor suppressor gene, PTEN (phosphatase and tensin homolog, GeneID: 5728) revealed the existence of a longer protein isoform resulting from use of an alternative in-frame upstream CUG initiation codon found at the center of a palindromic sequence upstream of the canonical mRNA translation start codon (13). Strong experimental data indicated that this mitochondrial-specific isoform initiates with a leucine, rather than a methionine (14). The RefSeq data model for eukaryotes provides one transcript explicitly linked to one protein. Therefore, two identical transcript records were provided to reflect translation from the alternate initiation codons; NP_000305.3 represents the 403 amino acid protein that uses the canonical methionine start codon, while NP_001291646.2 represents the mitochondrial-localized 576 amino acid protein that initiates with a leucine. Thus, the curation process serves a dual purpose of providing accurate reference sequences that facilitate precise and reproducible genome annotation and providing records that include relevant biological information. In this section we discuss recent updates, improvements we have
