Accurate, scalable cohort variant calls using DeepVariant and GLnexus

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Population-scale sequenced cohorts are crucial for genetic research.
• Processing raw sequencing data into analysis-ready variant callsets is a significant challenge.

Purpose of the Study:

• To introduce and optimize an open-source cohort-calling method.
• To improve the quality and reduce the cost of generating cohort-level variant data.

Main Methods:

• Utilized DeepVariant for accurate variant calling and GLnexus for scalable merging.
• Optimized the method using quality metrics like variant recall, precision, and Mendelian consistency.
• Evaluated the pipeline on 1000 Genomes Project (1KGP) data.

Main Results:

• The developed method demonstrated consistent quality improvements over existing best practices.
• Achieved superior callset quality metrics and imputation reference panel performance compared to GATK Best Practices.
• The optimized pipeline showed reduced costs for cohort variant calling.

Conclusions:

• The open-source cohort-calling method provides a high-quality, cost-effective solution for genetic analyses.
• Publicly releasing 1KGP variant calls and the pipeline fosters further development in cohort merging and genetic variation studies.