Missing by design: why genomic metadata must be captured at submission, not recovered from the literature
Plant and metazoan genome assemblies are being deposited in the International Nucleotide Sequence Database Collaboration (INSDC) archives at an accelerating rate, but the biological context needed to interpret them, such as ploidy, cultivar, and subgenome structure, is often missing from the record: these attributes are optional in most submission checklists and, when supplied, are entered as free text. We describe the gap in three layers. Ploidy varies within a species and cannot be inferred from taxonomy; chromosome-naming conventions differ between assemblies of the same crop; and that inconsistency is concentrated in species named by subgenome and is not associated with the size of the research community, which points to the lack of a shared convention. As a case study, we built a pipeline that recovers these attributes from the publications linked to an assembly. On a benchmark of 82 assemblies, its current version reported one wrong ploidy, and across 1,059 plant assemblies an earlier, less strict version found the ploidy in a publication for about half; for nearly a third, it found a publication but no statement of ploidy in it. Recovery from the literature can annotate assemblies already deposited but cannot stop new ones from arriving incomplete. We therefore propose that ploidy, cultivar, and subgenome structure be captured at submission as mandatory, controlled-vocabulary attributes, building on existing INSDC and Genomic Standards Consortium checklists.