Cross-platform benchmarking of somatic BRCA1/2 detection: four HRD NGS workflows against the Myriad MyChoice CDx reference assay

Introduction:Somatic BRCA1/2 testing is a cornerstone of ovarian cancer (OC) management, guiding PARP inhibitor (PARPi) therapy and informing hereditary cancer risk. While myChoice CDx is widely adopted as a clinical reference, many laboratories implement alternative next generation sequencing (NGS) workflows with different bioinformatics pipelines. Real-world benchmarking is essential to assess concordance and identify clinically relevant divergences. Methods:In the MITO16A trial we analyzed 100 formalin-fixed paraffin-embedded (FFPE) OC samples processed across four NGS pipelines within the Italian MITO network, referred to here as Lab1 and Lab4. The molecular analyses were performed using one academic workflow (Agilent OneSeq/VarDict; Lab1) and three commercial platforms (OCAplus/Ion Reporter, SOPHiA DDM, and TSO500/DRAGEN; Lab4). BRCA1/2 variant calls from all pipelines were compared with the myChoice CDx reference assay. Variant classification [pathogenic/likely pathogenic (P/LP) vs VUS], QC metrics, and CNV detection were centrally curated. Concordance was assessed at the sample level (presence or absence of P/LP variants), and all discordant cases underwent manual review to identify technical drivers of divergence. Results:Two samples (2/100) failed myChoice testing and were excluded, leaving 98 evaluable cases. Complete concordance across all pipelines and the reference assay was achieved in 86 of 98 samples (87.8%). The remaining 12 discordant cases (12.2%, 12/98) clustered into four recurrent technical categories: (i) exon-level CNV detection limits; (ii) low-VAF variants impacted by FFPE-related DNA damage; (iii) divergent VUS reporting policies (notably, VUS not reported by TSO500/DRAGEN); and (iv) transcript/HGVS nomenclature inconsistencies. An illustrative case harboring dual BRCA1/2 pathogenic variants highlighted the influence of local coverage and filtering thresholds. No pipeline exhibited systematic under- or over-calling trends. Conclusion:Cross-platform concordance for BRCA1/2 classification was moderate, with a non-negligible proportion of discordant calls across academic and commercial NGS workflows. These findings underscore the need for greater standardization and harmonization of tumor BRCA testing to ensure consistent and clinically reliable reporting. Our results have clinical implication in the general practice in the selection of predictive BRCA variants for therapy and in the selection of patients to be evaluated for germline mutations in the screening for hereditary syndromes.