PARADIGM is a factor graph method for inferring biological pathway activity from gene expression and gene copy number sample data. More information about PARADIGM is available at:

• http://bioinformatics.oxfordjournals.org/content/26/12/i237
• http://bioinformatics.oxfordjournals.org/content/29/13/i62

• pathway file
  • Here at UCSC, we use our superpathway. The current version is superpathway_4.0.no_chemicals.tab. You could create your own pathway file by following the format described at https://dna.five3genomics.com/paradigm/help.
• evidence files (expression data & copy number data)
  • The sample data are provided in tab-delimited 2-D matrix format.
    • samples in the columns
    • genes in the rows
  • The gene names must match the gene names in the pathway file (usually HUGO symbol).
  • The sample names must be common across data files as well as in the same order.
    • use commonColumnOrder.py
  • Are the features (genes) required to be common across data files as well as in the same order?
    • use commonRowOrder.py
  • normalizing expression data
    • If no match normal data is available, normalize against the cohort itself using a gene-wise, mean-center method.
      • Rscript normalize.r mRNA_matrix.tab mean_shift row > mRNA_matrix.meanshifted.tab
    • If normal data is available, then normalize against the normal data.
      • (Do we have a script for doing this?)
  • Is there any normalization requirement for copy number data?

• starting parameters and training parameters?

Submit the job via website at http://paradigm.five3genomics.com/.

(What to do with the results?)

(modulation filter)

(constituent pathways most enriched for changed activity level)