False Novelty from Taxid Retagging in Historical NCBI Taxonomy Reconstruction

Abstract

Resolving a sequence's identifier via a current-day lookup manufactures false novelty in historical reference-database reconstruction whenever the identifier has been renumbered — a known, tooled mechanism whose rate and consequence had not been quantified. Among 56 sequences flagged as newly created between two snapshots, only 17 were genuinely absent; 39 existed under a predecessor identifier merged into today's. Retagging these 39 recovered 99.3% correct classification, versus 0.0% under naive tagging. Library-wide, this recovers 218-269 identifiers per snapshot for ten of thirteen transitions studied. Our correction is a lower bound, missing deletion and reuse. We give five recommendations to avoid it.