Bipartite Graph-Based Hierarchical Clustering

Bipartite graph-based hierarchical clustering, developed for pharmacogenomic datasets and datasets sharing the same data structure. The goal is to construct a hierarchical clustering of groups of samples based on association patterns between two sets of variables. In the context of pharmacogenomic datasets, the samples are cell lines, and the two sets of variables are typically expression levels and drug sensitivity values. For this method, sparse canonical correlation analysis from Lee, W., Lee, D., Lee, Y. and Pawitan, Y. (2011) is first applied to extract association patterns for each group of samples. Then, a nuclear norm-based dissimilarity measure is used to construct a dissimilarity matrix between groups based on the extracted associations. Finally, hierarchical clustering is applied.