More than 2,500 coding genes in the human reference gene set still have unsettled status.

Area of Science:

• Genomics
• Proteomics
• Bioinformatics

Background:

• The human proteome annotation relies on major databases like Ensembl/GENCODE, RefSeq, and UniProtKB.
• Previous analysis in 2018 revealed significant disagreements among these databases regarding the coding status of human genes.

Purpose of the Study:

• To re-evaluate the consistency of human proteome annotation across Ensembl/GENCODE, RefSeq, and UniProtKB using updated datasets.
• To identify and analyze genes with potential non-coding features within the annotated coding gene sets.

Main Methods:

• Comparative analysis of updated gene annotations from Ensembl/GENCODE, RefSeq, and UniProtKB.
• Examination of 21,873 coding genes using eight features indicative of non-coding gene characteristics.
• Exclusion of read-through and immunoglobulin fragment genes for a refined comparison.

Main Results:

• Despite updates, significant discrepancies persist, with 2,606 annotated genes having disputed coding status.
• Over 700 genes were merged or reclassified between the 2018 and current analyses.
• More than 2,000 genes annotated as coding exhibited features suggesting they may be non-coding or pseudogenes.

Conclusions:

• Human proteome annotation databases still exhibit substantial disagreement on gene coding status.
• A considerable number of genes presumed to be coding may actually be non-coding or pseudogenes.
• Current annotation practices likely overestimate the true number of protein-coding genes.