ChosenHRDw: A novel tool for the detection of homologous recombination deficiency(HRD) using low-pass whole-genome sequencing.

e17573 Background: Homologous recombination deficiency(HRD) as a promising biomarker holds predictive and prognostic value in anticancer therapies, especially in ovarian cancer. Although next-generation sequencing techniques such as whole-exome sequencing (WES) and targeted sequencing have proven to be powerful detection of HRD, the latest low-pass (1x) whole-genome sequencing (WGS) with characteristic large-scale patterns for HRD detection status, as a cost-effective screening strategy, have established feasibility. In this study, we developed the ChosenHRDw, a comprehensive algorithm, which utilizing low-pass WGS to effectively classify the HRD status by defining the high or low HRD score. Methods: In this work, a correction method base on GC-content was applied. Then we took a window size of 1Mb resolution, and removed the error alignments genome bins basing on the healthy cohort data. Baseline of each bins was constructed from the data of 100 healthy volunteers. We characterized and quantified 8 HRD-related signatures: LOH score, TAI score, LST score, chromosomal instability index(CIN index), wGII(weighted genome instability index), CNV ratio (copy number ratio in bins), HRR index (HRR gene instability index), TFBS ratio (copy number ratio in transcription factor binding sites) as the variables in a random forest model. Results: A total of 140 patients (pts) were enrolled in the discovery cohort, which including 70 high HRD score and 70 low HRD pts were confirmed by HRDetect on paired tumor/normal samples via 1123plus-genes panel targeted sequencing. The AUC was 0.93. Independent validation was conducted in a validation cohort of 60 pts, which HRD high or low pts were 30 and 30, respectively. The sensitivity = 0.87 and specificity = 0.9. Conclusions: We classified HRD into high vs low using ChosenHRDw, which showing strong concordance. Our analyses demonstrated that the prediction of HRD status can be achieved from low-pass WGS. Due to the limitation of relatively small sample size, real-world studies with a larger number of patients are needed to verify our algorithm in the future.