Collaborative research with National Cancer Institute in the US published in Cell
Prof. LEE, JOO SANG
Dasol Kim, Youngmin Chung
Professor Joo Sang Lee’s research team (Next-Gen Medicine Lab, Department of Artificial Intelligence and School of Medicine) demonstrated that they can identify which therapies may be particularly beneficial for individual patients by only looking at the patients’ molecular markup before treatment. When applied to data from a wide panel of different cancer targeted and immunotherapy clinical trials, the approach termed SELECT was successfully predictive of patient responses to these therapies in about 80 percent of the trials. These findings were reported April 13, 2021, in Cell.
The study is based on identifying synthetic lethal interactions – a functional interaction between two genes whose co-inactivation leads to cancer cell death. Over 20 years ago, synthetic lethality has been proposed to have a great potential to revolutionize cancer treatment. This is partly because these interactions provide an opportunity to selectively kill only tumor cells while sparing normal cells by targeting synthetic lethal pairs of specific genes inactivated in a tumor. It has been investigated as a means to treat cancer, with some specific treatment regimens already used in the clinic. However, it is thought that many such treatment opportunities remain to be discovered and SELECT offers a computational method for identifying such treatment options for individual patients. By analyzing tumor transcriptomics data, this approach can identify actionable tumor vulnerabilities that are not readily evident by traditional mutational- and gene fusion-based sequencing approaches.
In collaboration with Eytan Ruppin (Cancer Data Science Laboratory at National Cancer Institute), Joo Sang Lee together with his students Youngmin Chung and Dasol Kim has been leading the development of computational tools to identify synthetic lethality. In this most recent study, they assembled a broad collection of 35 published transcriptomic datasets from targeted and immunotherapy cancer clinical trials across 10 different cancer types. They applied SELECT to predict the treatment response of the patients, given their tumor molecular data, finding the SELECT signatures to be highly accurate in 80 percent of the trials.
“To the best of our knowledge, SELECT is the first approach that systematically achieves these moderate, but helpful levels of accuracy across many different therapies and cancer types,” Lee says. Further investigation is now underway in collaboration with several clinical teams at the NIH Clinical Center to bring SELECT into clinic. The team hopes these prospective studies will further improve SELECT in the next few years, and if successful, establish SELECT as a complementary precision oncology approach for enhancing cancer-patient care.