Integrative Clustering Strategy and Subtype Analysis in Glioblastomas

The NCI/NHGRI initiated Cancer Genome Atlas (TCGA) project is a coordinated effort to catalogue the entire spectrum of genomic, epigenomic and transcriptomic alterations in the cancer genome. The TCGA network has recently published an interim analysis of DNA copy number, gene expression, DNA methylation aberrations, and mutation profiles of a selected set of cancer genes in 206 glioblastomas. Integrative analysis of multiple high-dimensional genomic data types is a new challenge. Major statistical issues that arise include generalizing dimension reduction to multiple high-dimensional correlated data sets, adjusting for spatial correlation along genomic positions, and modeling the within- and between-data type covariance structures. In this talk, I will introduce a lasso-regularized joint latent variable modeling framework to deal with these issues. Preliminary results of a subtype analysis in the GBM data set will be presented.