High Accuracy Small Variant Calling Using Machine Learning

ASHG 2022, Los Angeles, CA

Sentieon DNAscope: high accuracy small variant calling using machine learning supporting short read platforms.

Brendan Gallagher, Donald Freed, Zhipan Li
Sentieon Inc, San Jose, CA 95134

We present DNAscope, an accurate and efficient short-read germline small-variant caller. DNAscope combines the robust and well-established preprocessing and assembly mathematics of the GATK's HaplotypeCaller with a machine-learned genotyping model. Benchmarks of DNAscope and DNAseq (Sentieon's GATK-matching germline variant calling pipeline) demonstrate that DNAscope achieves superior SNP and insertion/deletion accuracy with reduced computational cost. DNAscope's machine-learning genotyping model is tuned for established and newly developed short read sequencing platforms including Illumina, Element Biosciences, and Ultima Genomics.

Download Abstract