A0798
Title: Dependent spike-and-slab for drug combinations in diabetic kidney disease
Authors: Jiefeng Bi - University of Milano-Bicocca (Italy) [presenting]
Abstract: Diabetic kidney disease (DKD) is a serious long-term complication of type II diabetes mellitus, primarily resulting from glucose metabolism disturbances associated with insulin resistance. Bayesian augmented learning (BAL) is employed, which leverages information across stages via dependent spike-and-slab priors to enhance treatment adaptation. False discovery rate (FDR) is also introduced as a method to identify the most significant variables at each intervention stage. This approach is applied to data from the PROVALID study, a prospective observational cohort of type II diabetes mellitus patients designed for biomarker validation. To adapt BAL for the PROVALID dataset, the original two-stage framework is extended to a four-stage model. Furthermore, a Bayesian predictive model is developed to determine optimal drug combinations for improving prognosis in individual DKD patients. The approach successfully identifies key molecular biomarkers and clinical parameters at each treatment stage, specifically in relation to drug effects on estimated glomerular filtration rate (eGFR).
