Cell-Level Auditing of an Orthology-Database-Derived Gene Presence/Absence Matrix Reveals Lineage-Specific Gene Gain, Duplication, and Pseudogenization

Abstract Cross-species gene presence/absence (PA) matrices are widely used to infer metabolic-pathway evolution, but orthology database clusters may contain deeply diverged homologs, lineage-specific duplicates, or pseudogenes, leading to erroneous assignments. Using OrthoDB v12 as the initial data source, we constructed a PA matrix comprising 34 human genes involved in nitrogen waste excretion and ammonia metabolism across 60 species, totaling 2,040 gene-by-species cells. Cells with uncertain assignments or distributions inconsistent with biological expectations were iteratively audited using NCBI/RefSeq searches, whole-genome tblastn, competitive paralog discrimination, locus-level clustering, phylogenetic priors, and published evidence. The final matrix contained 412 cells without support for the focal ortholog, 1,578 OrthoDB-only presence calls, 38 presence calls additionally supported by RefSeq or tblastn, 6 ancestral or unresolved homolog calls, 5 known pseudogenes, and 1 cluster bleed-through call. Sequence-level assessment of 20 randomly sampled OrthoDB-only presence cells yielded confirmation rates of 75% under a strict single-HSP criterion and 85% under a locus-aggregate criterion. Direct sequence evidence supported all four highest-risk absence assignments. All six candidate GLUD loci examined in non-ape vertebrates favored GLUD1 over GLUD2, with five exceeding the 2σ discrimination threshold, supporting the ape-specific retrocopy origin of GLUD2. Two loci on different chromosomes in Anolis carolinensis showed GLUD1-favored signals, suggesting a candidate GLUD1 duplication in this lineage. UOX sequences from four great apes shared a codon 33 stop mutation, supporting UOX pseudogenization in the Hominidae ancestor. These findings demonstrate that database cluster membership cannot be equated directly with focal-ortholog or functional-gene presence. Cell-level auditing provides a transparent and traceable framework for interpreting orthology-database-derived PA matrices and lineage-specific gene gain, duplication, and pseudogenization.