Abstract C018: Disparity subtyping: Bringing precision medicine closer to disparity science

The genomics revolution also spawned the dawn of precision medicine. As in the National Research Council definition, if its promise is fully realized, then more accurate decisions about individual patient treatment decisions and outcomes will be possible. Disparities researchers have also begun looking to the precision medicine paradigm with the hope that some incorporation of its principles will allow for a more focused and precise path forward to reduce population disparities. While the emphasis may switch to populations from individuals, central to the paradigm still is the ability to classify individuals into subpopulations who differ in meaningful ways with respect to underlying biology and outcomes. Identification of these subpopulations is an active area of precision medicine research. For instance, there are countless papers on molecular subtyping of various cancer phenotypes. How to do such a thing in disparity science has proven elusive since it requires identifying disparity subpopulations, which is a somewhat abstract concept. In this paper we present two different strategies—level set identification and peeling. The former is based on a recursive partitioning algorithm combined with clustering of similar partitions; the latter adopts a strategy of sequentially searching for and then extracting extreme difference subgroups in a population. Using series of simulation studies and then also studying various cancer outcomes from The Cancer Genome Atlas (TCGA) repository, we demonstrate that such disparity subtypes can indeed be found, characterized, and then validated on test data.

Citation Format: J. Sunil Rao, Huilin Yu, Jean-Eudes Dazard. Disparity subtyping: Bringing precision medicine closer to disparity science [abstract]. In: Proceedings of the Eleventh AACR Conference on the Science of Cancer Health Disparities in Racial/Ethnic Minorities and the Medically Underserved; 2018 Nov 2-5; New Orleans, LA. Philadelphia (PA): AACR; Cancer Epidemiol Biomarkers Prev 2020;29(6 Suppl):Abstract nr C018.