Weighted sparse PCA
This github repository contains all matlab and R code and functions that support the weighted sparse PCA manuscript. Also included are (derived) data and results. The repository contains a MATLAB branch for all matlab code, a R branch for all R code, a DATA branch with data that were derived from public data, and a RESULTS branch with the results obtained on simulated and empirical data.
-Gene expression data: Run the R scripts R/SCRIPT_ProcessingGSE29614_2007.R and R/SCRIPT_ProcessingGSE29617_2008.R. This retrieves the expression data from the online database and creates RMA pre-processed expression matrices for both seasons. The resulting pre-processed gene expression data are stored in DATA/GSE29614_rma.txt (2008 season) and DATA/GSE29617_rma.txt (2007 season). Besides, phenotype data with subject ids and the number of days after vaccination are created (DATA/GSE29614_pdata.txt and DATA/GSE29617_pdata.txt) along with a table containing the variance parameters in the two-component model resulting from the GESTr package (DATA/GSE29614__RLparsD0.txt and DATA/GSE29617__RLparsD0.txt).
-Antibody titers: These can be created also from the R scripts R/SCRIPT_ProcessingGSE29614_2007.R and R/SCRIPT_ProcessingGSE29617_2008.R. In these scripts the antibody titers are extracted from the CEL files and written to DATA/GSE29614_titers.txt (2008 season) and DATA/GSE29617_titers.txt (2007 season). From these files, the files DATA/TIVtiters_2007.txt and DATA/TIVtiters_2008.txt have been created by removing the textual descriptions and only retaining the numerical values. These files contain the antibody titers for three influenza strains and measured just before and 28 days after vaccination
Because we use the baseline-corrected gene expression data obtained three days after vaccination, the day 0 and day 3 data have to be matched on their subject identifiers and the difference scores have to be calculated. Note that the analyses are on the raw data (no centering nor scaling to unit variance). To reproduce these steps, run the following MATLAB scripts.
a. 2007 sample
- MATLAB/Script_Preprocessing_2007.m creates DATA/DATA2007.mat, DATA/DATA_D0_2007.mat, DATA/DATA_D3_2007.mat, and TIVTITER2007.txt
b. 2008 sample
- MATLAB/Script_Preprocessing_2008.m creates DATA/DATA2008.mat, DATA/DATA_D0_2008.mat, DATA/DATA_D3_2008.mat, and TIVTITER2008.txt
- Unweighed sparse PCA using the R PMA package: R/script_pma_influenzadata_FINAL.R
- Ordinary PCA, weighted PCA, and weighted sparse PCA using our WSPCA method implemented in MATLAB:
MATLAB/script_AnalysisInfluenzaData.m: this script requires two external matlab functions: fig.m and exportfig.m Available from: http://www.mathworks.com/matlabcentral/fileexchange/30736 and https://nl.mathworks.com/matlabcentral/fileexchange/727-exportfig
Annotate the probe-sets with non-zero weights resulting from WSPCA: R/Annotation.r
Generation of the data: MATLAB/Simulation_GenerateData.m
Analysis of the data: MATLAB/Simulation_AnalyzeData.m for PCA, WPCA, and WSPCA, and R/Simulation_AnalyzeData.r for SPCA
Creation of the boxplots: R/script_wspca_pma_boxplots_final_Prandom.r