The Platinum Pedigree: a long-read benchmark for genetic variants

Area of Science:

• Genomics
• Human Genetics
• Bioinformatics

Background:

• Genome sequencing advances have improved variant calling, but performance quantification is challenging due to a focus on specificity over completeness in complex regions.
• Existing truth sets are insufficient for evaluating variant calling in difficult-to-analyze genomic areas.

Purpose of the Study:

• To develop a more comprehensive human genome variant truth set.
• To improve the evaluation of variant calling performance across different sequencing platforms.
• To establish high-confidence truth sets for small variants, tandem repeats, and structural variants.

Main Methods:

• Utilized Mendelian inheritance principles within a large pedigree (CEPH-1463) to filter variants.
• Applied variant calling across PacBio HiFi, Illumina, and Oxford Nanopore Technologies sequencing platforms.
• Generated a comprehensive variant map covering 2.77 Gb of the GRCh38 human genome assembly.

Main Results:

• Created a variant map with over 4.7 million single-nucleotide variants, 767,795 insertions/deletions (indels), 537,486 tandem repeats, and 24,315 structural variants.
• Added approximately 200 Mb of high-confidence regions to the human genome reference, including an 8% increase in small variants.
• Developed the first tandem repeat and structural variant truth sets for the NA12878 family.

Conclusions:

• The developed comprehensive truth set enhances variant calling evaluation, especially in complex genomic regions.
• Retraining the DeepVariant model with this new data reduced genotyping errors by approximately 34%.
• This work provides a valuable benchmark for improving genome sequencing and variant analysis tools.