All Times EDT
Keywords: unsupervised machine learning technique, CART regression, Stratified-Medicine
In early drug discovery, scientists focus on initial research to better understand a disease or clinical condition with unmet medical need. Statisticians play a big role in helping the scientist to understand the relationship of a specific biomarker or therapeutic target with the disease of interest. We recently assessed whether a circulating hormone, proinsulin, could be used to identify subgroups of patients with type 2 diabetes mellitus, who might be helped by different therapeutic approaches. When clinically defined subgroups are known, supervised cluster analysis can be used to understand the relationship between biomarker or target and subgroup. However, clear subgroups are not apparent in type 2 diabetes. In this presentation, we will discuss how an application of unsupervised machine learning technique was used to identify novel subgroups of patients with type 2 diabetes. More specifically, application of CART regression and Patient Response Identifier for Stratified Medicine (PRISM) algorithm, a flexible and powerful subgroup identification framework, proposed by Jemielita and Mehrotra will be used.