Whole-genome benchmarking reveals context-specific error rates in the Ultima UG100 and Illumina NovaSeqX Platforms.

Abstract

Whole genome sequencing (WGS) is becoming more common in research and clinical applications, but there is a paucity of comparative data between high-throughput WGS platforms. We benchmarked the Ultima Genomics UG100 system against Illumina NovaSeqX using the Genome in a Bottle (GIAB) WGS reference set sample HG002. Across NIST v4.2.1 benchmark regions, UG100 exhibited higher total variant-calling errors (27x), primarily driven by indel false negatives. Restricting to Ultima Genomics high confidence regions (UG HCR v3.1, 90.3% of GRCh38) reduced the error burden by 89.6%, indicating most errors lie outside these regions. Strong degradation of variant calling performances occurred in homopolymer tracts > 10 bp, and base calling error rates increased sharply after position 200 bp within each read, whereas Illuminas error rates were higher earlier in the read. Coverage dropouts in UG100 data were pronounced in GC-rich regions. Of note, 2.24% of ClinVar Pathogenic and Likely Pathogenic variants and 22.6% of a common catalogue of polymorphic STRs were found to be excluded from UG HCR v3.1, indicating potential impact to clinical applications. These results highlight context-specific genotyping errors of the UG100 and NovaSeqX platforms, and underscore the importance of whole genome benchmarking for adjudicating accuracy of NGS platforms beyond the current HG002 reference set.