Regularization Approach to Screening and Selection of Biomarkers
Yoonkyung Lee
Department of Statistics
Ohio State University
Friday, November 30, 2007, 12:30–1:30 pm
GEMS classroom, 3rd Floor in
Shriner's Building
Coffee, tea, and cookies will be provided
Abstract
In medical studies involving such genomic data as gene expression levels
and SNPs, a large number of potential biomarkers are available for
prognosis or diagnosis of a disease. Finding a subset of relevant
biomarkers is important for understanding of the disease and construction
of effective prognostic or diagnostic rules. The method of regularization
in the form of convex optimization is considered as a computationally
viable alternative to the combinatorial search of the relevant biomarkers.
In this talk, we present a fast and efficient linear programming
algorithm for tracking the path of varying subsets of biomarkers in the
regularization framework.
It allows a complete exploration of the vast space of composite
biomarkers and facilitates screening and selection. The algorithm will be
illustrated with numerical examples.