Regularized Longitudinal Regression to Detect Biomarkers for Nephrotic Syndromes

• Activity Number: 49
• Type: Invited
• Date/Time: Sunday, August 3, 2014 : 4:00 PM to 5:50 PM
• Sponsor: ENAR
• Author(s): Peter Song*+
• Companies: University of Michigan
• Keywords: high-dimensional ; nephrology ; nonparametric regression ; regularization ; sparsity

Abstract

Our preliminary data analysis experience with the NEPTUNE consortium has suggested that some molecular biomarkers may be associated with longitudinal renal function measures such as estimated glomerular filtration rate (eGFR) in a nonlinear fashion. To detect important biomarkers predicting nephrotic syndromes, we develop a regularized nonparametric mixed-effects model to derive a prediction model of longitudinal eGFR over a large number of molecular biomarkers. The proposed regularization method assists us to detect and evaluate sparse molecular signals. The novelty of our method is that it can determine automatically which biomarkers are unassociated, linearly associated, or nonlinearly associated with longitudinal eGFR. We will illustrate our method on both simulation studies and a data analysis.