Undeclared substitutions in PDB entries do not track source organism; reference curation does
Year: 2026
Type: preprint
Abstract: Undeclared sequence differences between a PDB entry and its UniProt reference do not separate cleanly by source organism, but the ability to audit an entry does. We scanned 71,652 polymer entities from human, mouse, rat, cow, pig, dog, cat and horse entries, aligning each deposited sequence to its SIFTS-mapped UniProt record and, where that record is unreviewed (TrEMBL), to a reviewed member of its UniRef cluster. The 44,607 entities from each entry's own organism with a UniProt mapping form the comparison set. Among wildtype-labeled entities, the rate of point-shaped undeclared substitutions was 12.2% in human and 13.2% in mouse (Fisher p = 0.13); cow, pig, dog, cat and horse were lower, and rat was higher (16.7%). Whole-sequence mismatch against the organism's own reviewed record was higher in rat (6.3% vs 3.3% in human) and mouse (4.8%), and lower in cow, pig, dog, cat and horse. For 70% of pig entities, 37% of dog, 20% of cat and 5% of horse, there is no same-species reviewed reference to check against. The organism effect in PDB annotation lies in the availability of a curated reference, not in a higher rate of deposition error.
Author: Varman, Michael T.
Language: English
Subfield: Structural Biology
Domain: Life Sciences
Open Access status: green
OpenAlex ID: W7225982831
