Skip to main page content
Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2017 Apr;23(4):517-525.
doi: 10.1038/nm.4292. Epub 2017 Mar 13.

HRDetect Is a Predictor of BRCA1 and BRCA2 Deficiency Based on Mutational Signatures

Free PMC article

HRDetect Is a Predictor of BRCA1 and BRCA2 Deficiency Based on Mutational Signatures

Helen Davies et al. Nat Med. .
Free PMC article


Approximately 1-5% of breast cancers are attributed to inherited mutations in BRCA1 or BRCA2 and are selectively sensitive to poly(ADP-ribose) polymerase (PARP) inhibitors. In other cancer types, germline and/or somatic mutations in BRCA1 and/or BRCA2 (BRCA1/BRCA2) also confer selective sensitivity to PARP inhibitors. Thus, assays to detect BRCA1/BRCA2-deficient tumors have been sought. Recently, somatic substitution, insertion/deletion and rearrangement patterns, or 'mutational signatures', were associated with BRCA1/BRCA2 dysfunction. Herein we used a lasso logistic regression model to identify six distinguishing mutational signatures predictive of BRCA1/BRCA2 deficiency. A weighted model called HRDetect was developed to accurately detect BRCA1/BRCA2-deficient samples. HRDetect identifies BRCA1/BRCA2-deficient tumors with 98.7% sensitivity (area under the curve (AUC) = 0.98). Application of this model in a cohort of 560 individuals with breast cancer, of whom 22 were known to carry a germline BRCA1 or BRCA2 mutation, allowed us to identify an additional 22 tumors with somatic loss of BRCA1 or BRCA2 and 47 tumors with functional BRCA1/BRCA2 deficiency where no mutation was detected. We validated HRDetect on independent cohorts of breast, ovarian and pancreatic cancers and demonstrated its efficacy in alternative sequencing strategies. Integrating all of the classes of mutational signatures thus reveals a larger proportion of individuals with breast cancer harboring BRCA1/BRCA2 deficiency (up to 22%) than hitherto appreciated (∼1-5%) who could have selective therapeutic sensitivity to PARP inhibition.

Conflict of interest statement

Competing Financial Interests

H. Davies, D. Glodzik and S. Nik-Zainal are inventors on a patent application encompassing the code and intellectual principle on this algorithm. A. Tutt has been in receipt of payments from the Institute of Cancer Research Rewards to inventors scheme associated with the invention of PARP inhibitors as therapy for BRCA1 and BRCA2 mutation associated cancers.


Figure 1
Figure 1. Whole genome profiling depicts differences between patients with BRCA1/BRCA2 mutated tumours and sporadic tumours.
Examples of genome plots for a typical sporadic breast cancer (left), a BRCA1 germline null (middle) and a BRCA2 germline null tumour (right). Features depicted in circos plots from outermost rings heading inwards: Karyotypic ideogram outermost. Base substitutions next, plotted as rainfall plots (log10 intermutation distance on radial axis, dot colours: blue, C>A; black, C>G; red, C>T; grey, T>A; green, T>C; pink, T>G). Ring with short green lines, insertions; ring with short red lines, deletions. Major copy number allele (green, gain) ring, minor copy number allele ring (red, loss), Central lines represent rearrangements (green, tandem duplications; red, deletions; blue, inversions; purple, interchromosomal events). Mutations in breast cancer driver genes are indicated around the circus plots. Below each circos plot are the histograms showing mutation counts for each mutation class; topmost histogram shows the number of mutations contributing to each substitution signature; middle histogram represents indel patterns; lowermost histogram shows the number of rearrangements contributing to each rearrangement signature.
Figure 2
Figure 2. Workflow for developing HRDetect
A. Workflow of the steps involved in the development of the HRDetect predictor

Initial training using 22 known germline BRCA1 and BRCA2 null samples.

Retraining using 77 BRCA1 and BRCA2 null samples to produce the final HRDetect predictor.

Validation on a further set of breast cancers and application to other data sets.

B. Box plots of the weights for the genomic features contributing to the HRDetect predictor. Range of values from 10 replicates of training in cross-validation, using 311 breast cancer samples; 77 BRCA1/BRCA2 null samples and 234 quiescent tumours. Red crosses indicate the final coefficients used in HRDetect.
Figure 3
Figure 3. HRDetect as a probabilistic classifier
A. Piechart depicting the BRCA1 and BRCA2 mutation status of samples in the 560 breast cancer set which produced, on the lefthand side HRDetect scores below the cut-off of 0.7 and on the right handside above 0.7. Purple = previously known germline BRCA1 and BRCA2 with loss of the alternative allele; blue = newly discovered gemline BRCA1 and BRCA2 with loss of the alternative allele; orange = somatic gemline BRCA1 and BRCA2 and DNA hypermethylation of BRCA1 promoter with loss of the alternative allele; black = monoallelic germline and somatic BRCA1 and BRCA2 retaining the alternative allele; grey = samples in which no BRCA1 and BRCA2 mutation has been detected. B. HRDetect scores for 560 breast cancer samples ordered lowest to highest scores from left to right. Coloured bars included both monoalleleic mutations and those with loss of alternative allele, purple; previously known germline BRCA1 and BRCA2 (24 in total of which 22 are biallelic and 2 mono-alleleic), blue; newly discovered germline BRCA1 and BRCA2 (36 in total of which 33 are bialleic and 3 monoalleleic), orange; somatic gemline BRCA1 and BRCA2 and hypermethylation of BRCA1 promoter (31 in total of which 22 are biallelic and 9 monoalleleic), black diamonds above the bars indicate monoallelic germline and somatic BRCA1 and BRCA2 retaining the alternative allele (14).
Figure 4
Figure 4. Performance of HRDetect and validation
A. ROC curves demonstrating the performance of HRDetect on 371 breast cancer samples as well as performance when simply using individual mutational signatures as a predictor of BRCA1/BRCA2 deficiency. B. Comparing the sensitivity of detection of BRCA1/BRCA2 deficient tumours across different types of sequencing experiments - high-coverage 30-40X genomes, low-coverage 10X genomes and whole exome sequencing (using HRDetect weights learned from WGS and retrained on WES data). 371 breast cancer samples are used in each case. C. Performance of HRDetect on other data sets. From left to right - a cohort of 80 new breast cancers, 73 ovarian cancers and 96 pancreatic cancers. Top panel shows ROC curves for each cancer type respectively. Bottom panel shows histogram of HRDetect scores. Blue = germline BRCA1 and BRCA2 mutations; orange = somatic BRCA1 and BRCA2 mutations; grey = no BRCA1 and BRCA2 mutation detected; black diamonds above the bars indicate mono-allelic germline and somatic BRCA1 and BRCA2 retaining the alternative allele.
Figure 5
Figure 5. Clinically relevant strengths of HRDetect
A. Genome plot from an FFPE sample from a patient with a germline BRCA1 mutation B. Contribution of mutation signatures, top, substitutions; middle, indels; bottom, rearrangements; and below representation of the HRDetect score. C. HRDetect scores for nine patients treated with neoadjuvant anthracyclines +/- taxanes. Duplicate pretreatment needle biopsy samples were available for five of the samples (Pre-treatment Biospy 1 and 2). One patient (PD9770) had multifocal tumours. One patient with extremely low tumour cellularity in both biopsies and with hardly any mutations, was excluded (PD9773). HRDetect scores are provided under each sample. Blue shading indicates patients with residual disease while patients shaded green had complete response to treatment.

Similar articles

See all similar articles

Cited by 114 articles

See all "Cited by" articles

MeSH terms