Optimizing an NGS low-pass–based method to detect genomic instability as a PARP inhibitor predictive biomarker in high-grade serous ovarian cancer.

Abstract

5567 Background: In high-grade serous ovarian cancer (HGSOC) PARP inhibitors constitute a standard treatment in tumors harboring Genomic Instability (GI). This biomarker constitutes a valuable predictive tool for the use of PARP inhibitors. Available commercial solutions to determine the GI status, present some caveats that must be overcome. This study aims to set up an academic test in a cost-effective manner by applying open-source R libraries to establish GI status in Formalin-Fixed and Paraffin-embedded (FFPE) samples. Methods: The study was carried out in two stages, technical and analytical setup and clinical validation. Firstly, 16 FFPE samples, 8 tumoral tissue from gynecological malignancies, and 8 from healthy tissues were sequenced. This step aimed to establish favorable sequencing conditions followed by the tune-in of specific parameters for the analytical pipelines, QDNA, and Shallow-HRD (R v.4.3.11). Secondly, 44 FFPE samples from patients diagnosed with HGSOC and known GI score (GIS) were used for clinical validation. For this analysis, GIS determined by GIInger from Sophia Genetic was considered the gold standard. The series was constituted of 23 samples carrying genomic instability (GIS>=0) and 21 stable samples (GIS<0). Of note, intermediate libraries from targeted sequencing performed in clinical routine were used as input for sequencing. Individual libraries were then pooled and sequenced in a NextSeq 2000 (2x100 paired-end) (Illumina, San Diego, CA, USA) to achieve a 0.5x coverage. Recalibration of the method was performed using Maxstat algorithm implemented in R. All the analyses were performed in Python v3.8 or R v.4.3.11. Results: Best technical and analytical results were obtained for the QDNA pipeline without an X chromosome and bin size of 1 mb plus shallowHRD. By using this parameter, clinical validation comparing the GI results with GIS score obtained from the Sophia Genetics test was performed. In-house determination of GI resulted in 23 samples classified as unstable (score>=20) and 21 stable samples (score < 20). Re-calibration of the score using 20 as a new cut-off to dichotomize the variable, showed an optimum performance with a concordance of 86.4 % (p < 6.2 -6) with GIS classification. Thus, borderline samples called by shallowHRD pipeline (5/44) were re-classified as stable. Both continuous scores showed a correlation of 0.855 (p < 1.5 -13) and an Area Under the ROC curve of 0.906, presenting an excellent performance as a biomarker for genomic instability. Conclusions: We present a validated, routine-based and cost-effective test to determine GI in HGSOC, being easily transferrable to daily practice.