Whole Genome Sequencing Variant Calling Accuracy Improvements by Sentieon DNAscope Pan-Genome Analysis Pipeline

The Sentieon pan-genome analysis workflow utilizes a pangenome graph to perform alignment and variant calling on short-read DNA sequencing data. By adopting a graph-structured reference representation, the workflow improves alignment accuracy and variant detection in complex genomic regions with high sequence diversity, compared to traditional linear-reference approaches.

Within the workflow, the vg toolkit is used for graph-based alignment and structural variant (SV) detection. An accelerated implementation of this component will be released in future updates. Sentieon software handles small variant (SNP/Indel) calling, CNV detection, tandem repeat variant analysis, data preprocessing, and comprehensive quality control reporting. The workflow can also integrate specialized genotyping modules for highly polymorphic regions such as HLA/KIR, as well as expanded repeat analyses.

Similar to the linear-reference DNAscope workflow, the pan-genome pipeline uses FASTQ sequencing reads as input and produces aligned BAM files, VCF variant calls, and detailed QC and runtime reports as final outputs, enabling seamless downstream research and clinical applications.

Accuracy Evaluation

Using the NIST v4.2.1 truth set for benchmarking, short-read sequencing data (HG002, 30× depth) processed with the Sentieon DNAscope PanGenome workflow showed a 1.8× reduction in total SNP + Indel errors compared to the linear-reference DNAscope workflow. Comparable accuracy improvements were also observed in datasets produced on other sequencing platforms. The proportional distribution of error types remained consistent, indicating that the performance gains are uniform across variant categories and genomic regions.

Benchmarking with the more challenging T2T-Q100 truth set also demonstrated clear accuracy gains from the pan-genome workflow, even though the absolute error counts increased due to the increased evaluation scope. Continued version improvements are expected to further enhance performance under T2T-Q100 conditions.

Expanding the evaluation to multiple reference standards (including HG001, HG003, and HG004) produced consistent results, with the pan-genome workflow demonstrating the strongest overall accuracy.

In the associated figure, values represent total SNP + Indel errors under the NIST evaluation framework, where lower error counts indicate higher accuracy.

Additional benchmarking results demonstrating these improvements were also presented at ASHG 2025.

Structural Variant Accuracy

The pan-genome workflow provides substantial gains in structural variant recall sensitivity, achieving performance approaching that of long-read sequencing workflows, particularly in regions with complex genomic structure.

Availability

The Sentieon pan-genome workflow is currently available in a public preview release.

Documentation is available at:

https://github.com/Sentieon/sentieon-cli/blob/main/docs/pangenome.md

Other classic versions of DNAscope, inlcuding DNAscope (short reads on linear genome), DNAscope-LongRead (PacBio and ONT reads), DNAscope-Hybrid (short and long reads integrated analysis) have documents at https://www.sentieon.com/support/.

About Sentieon

Sentieon develops highly optimized bioinformatics algorithms by applying deep expertise in modeling, machine learning, and high-performance computing to enable precision data for precision medicine.