Biclustering of Data Matrices in Systems Biology and Drug Discovery
This figure illustrates several of the applications for optimal methods that we have developed for clustering based on mixed-integer linear optimization (MILP). The left figure presents the separation of stress conditions and subsequent biclustering of metabolite concentration data, where metabolites of similar known function are shown to group together. The top-right figure illustrates the clustering of protein sequences for de novo protein design in order to assess important homology trends. The bottom-right figure demonstrates the utility of optimal clustering for analyzing drug inhibition data, where desirable target molecules of high percent inhibition were found to cluster in the upper left-hand corner of the data matrix.