Harrison Pielke-Lombardo, Rani Powers, Aik Choon Tan, James Costello
University of Colorado Anschutz Medical Campus
Queries GEO for all human gene expression studies that have .CEL files available. Saves a table of the studies and their descriptions to RData/full_human_expression_GEO.rds. NOTE: The GEOmetadb package seems to have trouble connecting to the GEO database now. Rani has a local copy of the database downloaded previously (5 GB) that was used to generate RData/full_human_expression_GEO.rds. The analysis can be started with the R/01-annotateStudies.R script.
Filters the big table of studies for those using Affymetrix Human Genome U133 Plus 2.0 Array with at least n = 6 samples. Annotate the studies for those that contain 'vehicle,' 'drug', 'treatment,' etc in the titles or descriptions. Write out table as a text file for manual annotation of control vs drug-treated samples.
Studies were manually annotated and saved in Data/GEO/human_GPL570_n6_toAnnotate_withMetaData.xlsx. Relevant columns from this table were saved to Data/GEO/human_GPL570_n6_toAnnotate_completed.txt for use in the next R script.
For each of the annotated studies in Data/GEO/human_GPL570_n6_toAnnotate_completed.txt, this script downloads the CEL files from GEO. A directory for each study is created inside Data/GEO/ and the CEL files are downloaded and unzipped.
Compiles the set of CEL files from each experiment and extracts them into a single expression matrix saves in Data/GEO/Expression_Matrices.
Creates a ranked differentially expression gene list for each treated-vs-control comparison in each study. Saves the .rnk files to Results/Tables/DEG_lists/rnk/.
Once GSEA has been performed on all of the ranked lists from the previous step, this script converts all of the xls files produced by GSEA into txt files to Results/GSEA_DSigDB/txt/. Also saves a final Comparison_Key.txt describing the studies.