Abstract Background: Homologous recombination deficiency (HRD), broadly defined as a loss of the cellular mechanism underlying homologous recombination, is often observed in breast cancer. HRD causes distinctive perturbations to tumor genomic architecture that allow for its molecular identification, while also rendering HRD+ cancers vulnerable to specific chemotherapeutic interventions. This makes the molecular identification of HRD a promising avenue in precision medicine of breast cancer. Specific features of HRD include the presence of large-scale transitions (LST), telomeric allelic imbalance (TAI) and Loss of Heterozygosity (LOH). Each are readily detectable via targeted next generation sequencing (tNGS) or via array-based genotyping. However, genome-wide approaches for HRD detection using cost-effective methods, such as low-pass sequencing (LP-WGS), remain relatively under-explored. Here, we investigated whether HRD signals can be successfully re-capitulated using LP-WGS technology and benchmarked our results against the current field standard (both tNGS and array genotyping). Methods: LP-WGS and tNGS was performed on 96 samples across a range of tumor types (including N=17 breast cancer samples). LP-WGS libraries were prepared using Nextera (Illumina) using 0.4ng DNA input, and sequenced to 0.5-1x coverage. tNGS libraries were prepared using TSO500 (Illumina) using 40-80ng input, and sequenced to >150x unique read coverage. Regions of CNV were estimated using CNVKit v0.9.6, and regions of LOH were estimated using a novel ancestry-aware method. Small variant detection was performed using the TSO500 v2.2.0.12 analysis pipeline. SNP array analysis of 12 tumor samples using Oncoscan (ThermoFisher) was also performed. CNV and LOH estimates derived from LP-WGS, TSO500 and SNP array data were calculated using Jaccard similarity, treating the SNP array data as the “ground truth”. Results: We benchmarked HRD signals derived from LP-WGS compared to the array-based calls and observed near perfect sensitivity for CNV gains across samples (Jaccard index=1.0), as well as for CNV losses between LP-WGS and SNP array (Jaccard index=1.0). We additionally noted that LPS-WGS calls captured both CNV loss and gains that were not detectable via the SNP array. For TAI, LP-WGS re-capitulated 7/10 unique signals also identified via array. We also observed high concordance between regions of the genome called LOH between both platforms (median Jaccard index=0.70, IQR=0.254), but noted an attenuation of sensitivity in samples where estimated tumor heterogeneity was high. We also evaluated LP-WGS CNV calls against the TSO500 assay and noted high sensitivity (96%; 94%) and specificity (89%; 91%) for both CNV gains and losses, respectively. Conclusions: Workflows incorporating LP-WGS can support the detection of HRD genome-wide, paving the way for a more affordable assay that may help to inform clinical decision making in the future treatment of breast cancer. Citation Format: Gillian Belbin, Jie An, Chase Mazur, Joseph Pickrell, Jeremy Li, Daniel Metzger, Shuang Gao, Erik Van Roey, Robert Seager, Sarabjot Pabla, Durga Prasad Dash, Jeffrey M. Conroy. Application of low-pass whole genome sequencing for the detection of Homologous Recombination Deficiency in breast cancer [abstract]. In: Proceedings of the 2022 San Antonio Breast Cancer Symposium; 2022 Dec 6-10; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2023;83(5 Suppl):Abstract nr P5-02-54.
Read full abstract