Accurate detection of sub-1% frequency somatic mutations by whole-genome sequencing

Continuous advances in read accuracy have positioned bulk whole-genome sequencing as a promising approach for somatic mutation discovery in research and clinical settings, but its utility for detecting low-frequency mutations in high-coverage data has not been thoroughly assessed. We utilized an experimental design featuring (1) primary human fibroblasts with naturally occurring mutations and expanded subclonal populations, (2) induced pluripotent stem cell (iPSC) lines corresponding to subclones, and (3) additional iPSC lines originating from other fibroblasts from the same donor, enabling comprehensive evaluation of mutation detection across analytical methods and sequencing depths. Mutations discovered in iPSC lines defined positive control mutations ranging from 0.1% to 25% variant allele fraction (VAF) and negative (i.e., absent) control mutations in fibroblasts. Using the negative controls, we determined PacBio and AVITI as the most accurate sequencing platforms in "easy" (i.e., reliably accessible by short reads) and "non-easy" genomic regions, respectively. AVITI was also the most accurate for indel sequencing. We benchmarked mutation detection in fibroblasts in "easy" regions using ∼1,360× Illumina short-read data for mutation discovery and combined ∼600× PacBio and AVITI data for confirmation. Our analyses demonstrated accurate and specific detection of point mutations below 1% VAF and non-homopolymer indels down to 1% VAF. Homopolymer indels could be reliably detected only at VAF >3%. Mutation detection was degraded in postmortem samples, for which we provide recommendations for additional filtering and data-quality assessment. Finally, we show that combining ∼200× coverage across tissues from a donor is a feasible strategy for detecting developmental mutations shared across tissues.