In the following descriptions, "dataname" is a place holder for one of
the three datasets - "moran", "scherzer" and "zhang".
0. install packages pipeGS fron cran (Optional if loading directly the
   pval_comparison_dataname.rda files provided)
1. place the files dataname_all.RData and computeCorr.R in the same
   folder, change directory into the folder, and run from shell
   "Rscript ./computeCorr.R dataname", where dataname is moran, scherzer
   or zhang.
2. The first time running the script for each dataset will take about
   20 minutes to compute phat.saddle1, phat1, phat2, phat3 for all 6180
   gene sets. It caches all the pvalues in the file
   pval_comparison_dataname.rda after the first run, so later runs
   automatically loads the corresponding images. In the end it
   prints the correlation Table 3 and the time Table 4 in the paper for Moran and Scherzer
   dataset.


Remark 1. dataname_all.RData contain the following objects
- X: the binary indicator for treatment and control groups with length n
- Y: the gene measurement matrix for all samples with size n X p,
  column names are labeled with gene symbols
- gs: defitions of gene sets with entrez.id
- entrezids.map: the mapping from gene symbol to entrez.id
- phat.mc: The monte carlo estimates for 6180 gene sets as described
  in the paper.

Remark 2. The gene set defitions are downloaded from
 http://software.broadinstitute.org/gsea/msigdb/collections.jsp
We used c2.all.v5.1.entrez.gmt and c5.all.v5.1.entrez.gmt. The current
 version is v6.1.

Remark 3. The latest "moran", "scherzer" and "zhang" data can obtained
   from the following urls respectively.
"https://ftp.ncbi.nlm.nih.gov/geo/series/GSE8nnn/GSE8397/matrix/GSE8397-GPL96_series_matrix.txt.gz",
"https://ftp.ncbi.nlm.nih.gov/geo/series/GSE6nnn/GSE6613/matrix/GSE6613_series_matrix.txt.gz",
"https://ftp.ncbi.nlm.nih.gov/geo/series/GSE20nnn/GSE20292/matrix/GSE20292_series_matrix.txt.gz"
These files, once downloaded can be read into R with "getGEO(filename =
file.path)" - getGEO is a function in bioconductor package GEOquery.
These data have changed since we did our first analysis - more control cases
have been added- hence we provide the old data in the format in the
above mentioned RData files.