A Benchmark of In-House Homologous Recombination Repair Deficiency Testing Solutions for High-Grade Serous Ovarian Cancer Diagnosis

Abstract

Homologous recombination deficiency (HRD) has become an important prognostic and predictive biomarker for patients with high-grade serous ovarian cancer who may benefit from poly-ADP ribose polymerase inhibitors (PARPi) and platinum-based therapies. HRD testing provides relevant information to personalize patients' treatment options and has been progressively incorporated into diagnostic laboratories. Here, we assessed the performance of an in-house HRD testing system deployable in a diagnostic clinical setting, comparing results from two commercially available next-generation sequencing (NGS)-based tumor tests (SOPHiA DDM^TM HRD Solution and AmoyDx^® (HRD Focus Panel)) with the reference assay from Myriad MyChoice^® (CDx). A total of 85 ovarian cancer samples were subject to HRD testing. An overall strong correlation was observed across the three assays evaluated, regardless of the different underlying methods employed to assess genomic instability, with the highest pairwise correlation between Myriad and SOPHiA (R = 0.87, p-value = 3.39 × 10^-19). The comparison of the assigned HRD status to the reference Myriad's test revealed a positive predictive value (PPV) and negative predictive value (NPV) of 90.9% and 96.3% for SOPHiA's test, while AmoyDx's test achieved 75% PPV and 100% NPV. This is the largest HRD testing evaluation using different methodologies and provides a clear picture of the robustness of NGS-based tests currently offered in the market. Our data shows that the implementation of in-house HRD testing in diagnostic laboratories is technically feasible and can be reliably performed with commercial assays. Also, the turnaround time is compatible with clinical needs, making it an ideal alternative to offer to a broader number of patients while maintaining high-quality standards at more accessible price tiers.

Keywords: BRCA1/2; DNA repair; biomarkers; homologous recombination deficiency (HRD); next generation sequencing (NGS); ovarian cancer; poly (ADP–ribose) (PARP) inhibitors (PARPi).

Figure 1

Comparison of predicted genomic instability (PGI) scores across different next-generation sequencing (NGS)-based tumor assays. (A) Pairwise correlation of PGI score using three commercially available HRD testing solutions: Myriad MyChoice CDx, SOPHiA DDM^TM HRD solution, and AmoyDx HRD Focus Panel. The highest pairwise correlation observed was between Myriad and SOPHiA (R = 0.87, p-value = 3.39 × 10⁻¹⁹). Bootstrapping-based 95% confidence interval of the regression estimate shown in gray translucent bands around the regression lines. (B) Observed discrepancies between the PGI score from Myriad MyChoice CDx and the results obtained from SOPHiA DDM^TM HRD solution and AmoyDx HRD Focus Panel. The number of samples that failed quality control (QC) metrics is in parentheses.

Figure 2

Correlation between BRCA1/2 status and predicted genomic instability (PGI) score association and consistency between the two biomarkers, BRCA1/2 status and genomic instability, in a subset of samples that were successfully executed in the three next generation sequencing (NGS)-based tumor assays and passed the quality control (QC) metrics. On top: Myriad MyChoice CDx; in the middle: SOPHiA DDM^TM HRD solution; and on the bottom: AmoyDx HRD Focus Panel. The SOPHiA test was able to detect one somatic pathogenic variant in BRCA1 with a variable allele fraction (VAF) of 9.8%, which was not previously reported by the other two tests. P/LP; pathogenic/likely pathogenic. Bars were ordered by PGI score within each assay, so they do not necessarily follow the same order between assays.

Figure 3

Determination of the precision performance of next-generation sequencing (NGS)-based tumor assays. Reproducibility of intra- and inter-run in-house HRD testing experiments for SOPHiA DDM^TM HRD solution and AmoyDx HRD Focus Panel.

Figure 4

Associations of pre-analytical variables with low-confidence next-generation sequencing (NGS) results. (A) Four pre-analytical variables evaluated between pass and low-confidence experiments: (i) estimated tumor content (ETC); (ii) functional DNA qPCR, (iii) DNA input; and (iv) DNA library yield. In our analysis, only functional DNA qPCR and DNA input were statistically significant (p-value = 3.52 × 10⁻⁴; p-value = 3.4 × 10⁻⁴). (B) Negative correlation between DNA library yield and residual noise (R = −0.62, p-value = 1.72 × 10⁻⁹). Residual noise is computed by measuring the standard deviation of the normalized genome sequencing coverage profile with respect to the smoothed normalized genome sequencing coverage profile. Bootstrapping-based 95% confidence interval of the regression estimate shown in gray translucent bands around the regression line. (C) Two examples of samples that have low (left) and high (right) estimated residual noise profiles.