Abstract
The 1000 Genomes Project is a foundational resource to modern human biomedicine, serving as a standard reference for human genetic variation. Recently, new versions of the 1000 Genomes Project dataset were released, expressed relative to the current version of the human reference sequence (GRCh38) and partially validated by benchmarking against reference truth sets from the Genome In A Bottle Consortium. We used our ultrafast genome comparison method (genome fingerprinting) to evaluate four versions of the 1000 Genomes Project datasets. These comparisons revealed several discrepancies in dataset membership, multiple cryptic relationships, overall changes in biallelic SNV counts, and more significant changes in SNV counts, heterozygosity and genotype concordance affecting a subset of the individuals. Based on these observations, we recommend performing global dataset comparisons, using genome fingerprints and other metrics, to supplement ‘best practice’ benchmarking relative to predefined truth sets.
Abbreviations used
- GIAB
- Genome In A Bottle Consortium
- GRCh37, GRCh38
- Genome Reference Consortium, human reference versions 37 and 38
- IGSR
- International Genome Sample Resource, http://www.internationalgenome.org
- SNV
- Single-nucleotide variant
- TGP
- Thousand Genomes Project
- TGP37, TGP37r, TGP38, TGP38L, TGP38r, TGP38S
- The TGP datasets studied in this work
