Tuesday, April 9, 2019
3:30 pm - 4:30 pm
701 Blockley Hall, 423 Guardian Drive, Philadelphia, PA 19104
Title: Considerations in the development of biologically informed treatment selection scores from high-dimensional omics data
Abstract: Precision medicine relies on matching biological and molecular characteristics of a patient or the patient's disease to therapy mechanism of action. The disease may develop or progress due to disruptions in the function of genes that code for proteins integral to the disease process. Therapies that can correct or replace the aberrant or missing gene products hold promise for the treatment of the corresponding disease. In oncology, many omics predictors based on multivariable scores have been developed for purposes of prognosis, and then secondarily assessed for whether they are useful also for choosing between different therapy options. Repurposing a prognostic predictor may be suboptimal because there are fundamental differences between the goals of prognostication and therapy selection. The modified covariates method of Tian and colleagues (JASA 2014;109:2350-2358) is one approach that has been proposed specifically for development of a therapy selection predictor. A biologically informed enhancement of the modified covariates approach is proposed in this talk. Performance of this biologically enhanced version is compared with that of the original modified covariates method on some real data from patients with cancer. In this context, several more general issues regarding appropriate evaluation of treatment selection predictor performance are highlighted.
Biometric Research Program, National Cancer Institute
Lisa M. McShane, Ming-Chung Li, George Wright, Ting Chen, Lori Long, Qian Xie, Jyothi Subramanian, Yingdong Zhao