Identifying Transcripts with Tandem Duplications from RNA-Sequencing Data to Predict BRCA1-Type Primary Breast Cancer

SIMPLE SUMMARY: Homologous recombination repair deficiency (HRD) is a biomarker for the response to PARP inhibitor anti-cancer treatment. Therefore, methods that detect the HRD phenotype in cancers in a (cost-)effective manner are pivotal. In this respect, the HRDetect and CHORD algorithms were developed to classify (the type of) HRD cancers from whole genome sequencing data. In addition, functional assays have also been established, but these require fresh cancer tissue. Here we present a novel method to specifically classify BRCA1-type HRD from RNA-sequencing data with high sensitivity. BRCA1-type cancers typically display small (<10 kb) tandem duplications, in contrast to BRCA2-type cancers. By detecting these small TDs among transcripts, we increase the toolbox for detecting HRD with a method that does not require whole genome sequencing of both tumor and normal tissue. ABSTRACT: Patients with cancers that are deficient for homologous recombination repair (HRD) may benefit from PARP inhibitor treatment. Therefore, methods that identify such cancers are crucial. Using whole genome sequencing data, specific genomic scars derived from somatic mutations and genomic rearrangements can identify HRD tumors, with only BRCA1-like HRD cancers profoundly displaying small (<10 kb) tandem duplications (TDs). In this manuscript we describe a method of detecting BRCA1-type HRD in breast cancer (BC) solely from RNA sequencing data by identifying TDs surfacing in transcribed genes. We find that the number of identified TDs (TD-score) is significantly higher in BRCA1-type vs. BRCA2-type BCs, or vs. HR-proficient BCs (p = 2.4 × 10(−6) and p = 2.7 × 10(−12), respectively). A TD-score ≥2 shows an 88.2% sensitivity (30 out of 34) to detect a BRCA1-type BC, with a specificity of 64.7% (143 out of 221). Pathway enrichment analyses showed genes implicated in cancer to be affected by TDs of which PTEN was found significantly more frequently affected by a TD in BRCA1-type BC. In conclusion, we here describe a novel method to identify TDs in transcripts and classify BRCA1-type BCs with high sensitivity.