This is the code for the second submission. The major changes from the first submission include Switch from DESeq to DESeq2 DESeq2 controls for the false positive rate differently, allowing for a more powerful statistical analysis. We switched all the analyses over to DESeq2. This dramatically changed the number of statistically significant genes. This was done in issue #9 for the code and issue #14 for the manuscript. We also upgraded to GRCh37.74 version of the human genome in issue #13. Counts Mapping Algoritm Based on utility, and avoiding the need to filter out extra transcripts that we are not analysing we mapped genes using HTSseq rather than Genomic Ranges. This was done in issue #1. Clinical Measurements We noted that some of the clinical measurements seemed to be non-normally distributed. We adjusted the clinical analysis script to first test for normality, then test for equal variance. The resulting p-values are now based on either a Wilcoxon, Welch or Student's T-Test. This was done in issue #16 (commit 6143ff1). Gene Set Enrichment Added analysis using GSEA to replace/complement the GOseq analysis. GSEA (http://www.broadinstitute.org/gsea/index.jsp) incorporates the rank of genes in the entire data set, rather than just the top significant hits. This allowed us to analyse more data sets including miRNA, TRANSFAC, Reactome. This was done in issue #12 and largely in commit 1e25f3b054631e549495c482eb5cbc655910c865. These data are presented in Supplementary Tables 2,3 and 4 as done in issue #20