This page provides a software implementation of the approach for team PRB (submission Team49) in the IMPROVER Inter-species Pathway Perturbation Prediction of the IMPROVER Species Translation Challenge.
Prerequisites:
-Pre-ranked GSEA from broad institute, file “gsea2-2.09.jar”. If a different version is used, then modify the file “gseapF.r” and specify the version you use.
– R statistical environment v 3.0.0 or higher on a system (linux like) that has the following packages installed:
– parallel
– limma
The entry point in this pipeline is the R script file: sc3prb.r. Download “gsea2-2.09.jar” from http://www.broadinstitute.org/gsea/login.jsp (you will need to provide an e-mail address) and place it in the root of the SC3 directory. The code reads the rat and human gene expression files for stimuli in the training set (“GEx_rat_train.txt”, “GEx_human_train.txt”) as well as the rat test dataset “GEx_rat_test.txt” and finds for each human gene the rat gene that mimics is behavior best. This homology mapping is saved as a named vector “respf” in a RData file named “respf.RData”. Names of the vector are human genes and values of the vector are rat genes. Then for each stimuli in the in the rat test dataset a moderated t-score is computed to rank genes and GSEA is called to obtain FDR value for each pathway. These are transformed in confidence levels that the pathway is activated and written in the file “PRB_SC3_predictions.txt” which follows the template provided by STC organizers ( “template_prediction_SC3_gene_sets_human.txt”).
To download the R script and required data sets click here.