Using Genomics to Prognosticate Survival in Ovarian Cancer
Updated: Oct 16, 2022
I am pleased to announce today that a passion project has finally been published in JAMA Network Open:
Graf RP, Eskander R, Brueggeman L, Stupack D. Association of Copy Number Variation Signature and Survival in Patients With Serous Ovarian Cancer. JAMA Network Open. 2021. PMID: 34181012
I had always wanted to do a project surrounding ovarian cancer; a member of my family’s life was shortened by the disease, and it remains a very difficult disease to treat.
The goal of precision medicine is to predict the future for individual patients. Preferably, to predict the outcomes of multiple routes a patient could take in their treatments, to enable choosing the route with the best outcome. From a data standpoint, three basic ingredients are needed for this recipe: biomarkers (i.e. genomics), what treatments they received, and what happened, i.e. the outcome. Typically, public databases don’t have all three, and if they do, there are not complete data on enough patients to make it work.
One day a few years back my PhD advisor Dwayne Stupack and I were catching up over a beer. The Cancer Genome Atlas (TCGA) database for ovarian cancer had recently been updated to have not only extensive genomic data (biomarkers), but survival data on almost 600 patients. Even more interesting, during the time period in which the patients were treated (mid 1990’s - 2010’s) the standard of care treatments were pretty uniform: surgery, then platinum+taxane chemotherapy. (While the standard of care has changed slightly since, this regimen remains the backbone, with some additional treatment added on top for some patients.) We could infer the third ingredient. The early pandemic lockdowns made this an attractive project to occupy my nights and weekends.
I won’t go into the details (the paper is open access here), but the short story is that we developed a genomic signature that could aid in the assessment of how long a patient might live given the foundational treatments for ovarian cancer. However, before this can be used clinically it needs to be validated in an external dataset, i.e. it needs to show consistent results to predict outcomes in a different set of patients.
Existing signatures for ovarian cancer make use of very complicated RNA data, and one strength of our work is demonstrating the potential to use much-more-simple-to-assess copy number variations in DNA toward this purpose. It has also identified new areas of the genome that might have relevance for future drug development, as alterations in these regions are associated with differential outcomes in these patients.
Thank you for dropping by!