Heat Map representing Human B-Cells analyzed using RNA CoMPASS.

Human transcript counts from the 45 B-cell samples were imported into the R software environment and analyzed using the edgeR package [15]. Genes with low transcript counts (less than 1 CPM (count per million)) in the majority of samples were filtered out. The Manhattan (L-1) distance matrix for the samples was computed using the remaining transcript counts, and this was taken as input for hierarchical clustering using the Ward algorithm. After assigning each sample to one of two groups identified by hierarchical clustering (Human B-Cell or Burkitt's Lymphoma), the glmFit function was used to fit the mean log(CPM) for each group and likelihood ratio tests were used to identify those genes that were differentially expressed, with adjusted P<0.05 following the Benjamini-Hochberg correction for multiple testing. The fitted log(CPM) values for the subset of genes that were differentially expressed in the LCL samples relative to the Burkitt's lymphoma samples were then clustered using the Euclidean distance and complete linkage algorithm to detect groups of co-expressed genes. The expression heat map displays the top 250 differentially expressed genes.