Ovarian cancer specific homologous recombination deficiency score

Our publication shows that ovarian cancers have a differential abundance of genomic lesions compared to other types of cancer.
Published in Cancer
Ovarian cancer specific homologous recombination deficiency score

Share this post

Choose a social network to share with, or copy the shortened URL to share elsewhere

This is a representation of how your post may appear on social media. The actual post will vary between social networks

The problem: which homologous recombination deficiency test should be used to guide the treatment of ovarian cancer with PARP inhibitors?

High-grade serous ovarian peritoneal or fallopian-tube cancer (HGSC) is the most common type of ovarian cancer. The 5-year overall survival of HGSC has remained poor, approximating less than 40%, with DNA-damaging therapy and cytoreductive surgery as the standard of care. However in recent years, the treatment of HGSC has been revolutionized by the use of PARP inhibitors.

Approximately half of HGSC have a deficiency in the homologous recombination DNA repair pathway (HRD). Cancer cells with HRD are unable to repair DNA double-strand breaks (DSB) and accumulate DNA lesions. HRD tumors are vulnerable to DSB-inducing agents such as platinum-based chemotherapy and PARP inhibitors. In clinical trials, it has been shown that patients with HRD tumors have excellent responses to PARP inhibitors as first-line maintenance treatment. Because of the response of HRD tumors to PARP inhibitors and considering the high cost of this treatment, it is critical to develop an accurate test for detecting HRD to guide the treatment of HGSC.

Why a specific HRD test for ovarian cancer?

The FDA-approved test MyriadMyChoise®CDx, which was used in clinical trials, quantifies three types of DNA lesions (loss of heterozygosity [LOH], large scale transitions [LST], and telomeric allelic imbalances [TAI]) enriched in BRCA1/2-mutant tumors. The criteria behind this test were developed using mixtures of ovarian cancer and breast cancer samples in small patient cohorts.

In our work, we showed that HGSC samples have a higher proportion of DNA lesions as compared to other types of cancers including breast cancers. With this in mind, we aimed to optimize the criteria for HRD specifically for HGSC using large multi-omics patient cohorts.

 Our optimization approach

For our optimization, we used the HGSC samples from the TCGA. Because there are other drivers for HRD beyond BRCA1/2 mutation, we selected HRD samples using a multi-omics approach: mutation, promoter hyper-methylation, and gene deletion status of BRCA1/2 or RAD51c-paralog genes. Then, instead of selecting the rest of the samples as non-HRD, we aimed to select samples with intact genes in the homologous recombination pathway (further called HRP samples). These HRD and HRP “ground truth” samples showed a distinct abundance of DNA lesions. Then, using an exhaustive statistical and machine learning approach, we identified the best selection criteria for LOH, LST, and TAI to distinguish HRD from HRP samples. Similarly, we used a bootstrapping approach to identify the number of selected DNA lesions that increase the accuracy of labeling HRD and HRP samples according to the ground truth.

Our optimization criteria, ovaHRDscar, led to a higher concordance with mutational signature 3, a genomic signature associated with HRD, as compared to the criteria used by MyriadMyChoise®CDx (Telli2016 criteria). Similarly, ovaHRDscar allowed us to distinguish our “ground truth” HRD and HRP tumors with greater accuracy than the Telli2016 criteria.

 Validation in external cohorts

Because HRD tumors are vulnerable to platinum-based chemotherapy, we tested if ovaHRDscar was able to identify the patients that got more benefit from platinum-based chemotherapy as compared to several other previously reported criteria. To do this, we used four independent validation cohorts. A clear improvement was observed in the validation cohorts PCAWG, DECIDER, and TERVA, where ovaHRDscar was able to detect the patients with longer progression free interval and overall survival as compared to the Telli2016 criteria. Modest improvement was observed in the TCGA cohort, which can be explained by that part of Telli2016 criteria was specifically optimized for outcomes in the TCGA cohort.

Optimization for triple-negative breast cancer

Finally, we replicated our optimization approach for triple-negative breast cancer (TNBC) samples in the TCGA (tnbcHRDscar). Similarly, tnbcHRDscar allowed us to distinguish our HRD and HRP tumors with greater accuracy than the Telli2016 criteria. In an external validation cohort, we observed that tnbcHRDscar allowed us to identify the patients with longer progression free interval and overall survival.

A cancer specific optimized test

Our results show that, on average, each type of cancer has a differential abundance of DNA lesions. To create better precision oncology, it might be important to create an HRD clinical test specific to each type of cancer. By using machine learning and exhaustive statistics, we identified the DNA-lesion patterns enriched in HRD samples. Our ovaHRDscar algorithm is freely available and has the potential to be used in clinics to guide the treatment of HGSC with PARP inhibitors.

Please sign in or register for FREE

If you are a registered user on Research Communities by Springer Nature, please sign in

Subscribe to the Topic

Cancer Biology
Life Sciences > Biological Sciences > Cancer Biology

Related Collections

With collections, you can get published faster and increase your visibility.

AI in precision oncology

This Collection is a partnership between npj Precision Oncology and npj Breast Cancer. It will bring together articles on all facets of AI in cancer research.

Publishing Model: Open Access

Deadline: Apr 19, 2024

Innovations in cancers of the central nervous system

This Collection invites research on tumors involving the central nervous system, including primary glial tumors, meningiomas and metastatic tumors in adults.

Publishing Model: Open Access

Deadline: Mar 01, 2024