89. Identification of therapeutic combinations for immune checkpoint inhibitors (ICIs) using explanatory subgroup discovery

Aly Abdelkareem

Olha Kholod

Olha Kholod is a PhD student in Biomedical Informatics at the University of Missouri-Columbia. Conjointly, she works as a Graduate Research Assistant in the Interdisciplinary Data Analytics and Search (iDAS) Research Lab under Dr. Shyu supervision. Olha was the recipient of the prestigious Fulbright U.S. Student Program Award from the U.S. Department of State in 2015. She graduated with Master of Science in Pathology degree from University of Missouri-Columbia in May 2017. Olha is a member of several professional and student organizations including the American Medical Informatics Association (AMIA) and International Society of Computational Biology (ISCB). Her current research is focusing on identification of immuno-targeted combination therapies using explanatory subgroup discovery for cancer patients. Outside of academics, Olha enjoys hiking, reading books and studying foreign languages.


Olha Kholod, William Basket, Danlu Liu, Jonathan Mitchem, Jussuf Kaifi, Chi-Ren Shyu

University of MO-Columbia, Columbia, MO, USA

Phenotypic and genotypic heterogeneity are characteristic features of cancer. This heterogeneity significantly limits therapeutic response and application, especially in patients without targetable mutations. One example of this is immune checkpoint inhibitors (ICIs), which represent one of the best therapeutic approaches to treat cancer. However, 50% of patients that are eligible for ICI therapy do not respond. Multiple methods for selection have been developed, but still lack the ability to consistently identify patients that will benefit from treatment. To address this problem, we developed an informatics framework that consists of two modules: subgroup discovery and drug target discovery. The subgroup discovery module identifies homogenous patient subgroups using phenotypic and genotypic parameters and explains the differences between subgroups using gene expression patterns. The drug targets discovery module employs a proportional odds model to identify targets for use in combination with ICIs. Focusing specifically on patients without targetable mutations, we hypothesize that most of these patients will benefit from compounds that have been used for immuno-targeted combination therapies. Our pipeline identified six specific drug targets and thirteen specific compounds for EGFR WT patients in four malignancies: head & neck cancer, lung adenocarcinoma, lung squamous carcinoma, and melanoma. Three out of six drug targets â?? FCGR2B, IGF1R and KIT â?? substantially increase the odds of having stable versus progressive disease, thus identifying both markers of poor prognosis and potential intervention. This approach may help to better select responders for combination therapy with ICIs and improve health outcome for cancer patients without targetable mutations.