Online Program

Coherent Multi-Task Feature Selection and Prediction from Pharmacogenomics Databases (303817)

Keywords: Random forest,Elastic net, Adaptive multi-task, Bayesian

Integrating multiple databases of similar tasks is a significant problem in biological data analysis. In this article we offer two methodologies for combining information across similar databases. First, we demonstrate that an adaptive multi-task elastic net for feature selection and Random Forest for prediction, can be used to borrow information across databases with superior predictive performance. Second, we offer a two stage Bayesian data fusion model to perform full predictive inference. We illustrate both methodologies with synthetic data and publicly available pharmacogenomics data.