discordantRun {discordant} | R Documentation |
Runs discordant algorithm on two vectors of correlation coefficients.
discordantRun(v1, v2, x, y = NULL, transform = TRUE, subsampling = FALSE, subSize = dim(x)[1], iter = 100, components = 3)
v1 |
Vector of Pearson correlation coefficients in group 1 |
v2 |
Vector of Pearson correlation coefficients in group 2 |
x |
ExpressionSet of -omics data |
y |
ExpressionSet of -omics data, induces dual -omics analysis |
transform |
If TRUE v1 and v2 will be Fisher transformed |
subsampling |
If TRUE subsampling will be run |
subSize |
Indicates how many feature pairs to be used for subsampling. Default is the feature size in x |
iter |
Number of iterations for subsampling. Default is 100 |
components |
Number of components in mixture model. |
The discordant algorithm is based on a Gaussian mixture model. If there are three components, correlation coefficients are clustered into negative correlations (-), positive correlations (+) and no correlation (0). If there are five components, then there are two more classes for very negative correlation (–) and very positive correlations (++). All possible combinations for these components are made into classes. If there are three components, there are 9 classes. If there are five components, there are 25 classes.
The posterior probabilities for each class are generated and outputted into the value probMatrix. The value probMatrix is a matrix where each column is a class and each row is a feature pair. The values discordPPVector and discordPPMatrix are the summed differential correlation posterior probability for each feature pair. The values classVector and classMatrix are the class with the highest posterior probability for each feature pair.
discordPPVector |
Vector of differentially correlated posterior probabilities. |
discordPPMatrix |
Matrix of differentially correlated posterior probabilities where rows and columns reflect features |
classVector |
Vector of classes that have the highest posterior probability |
classMatrix |
Matrix of classes that have hte highest posterior probability where rows and columns reflect features |
probMatrix |
Matrix of posterior probabilities where rows are each molecular feature pair and columns are nine different classes |
loglik |
Final log likelihood |
Charlotte Siska <siska.charlotte@gmail.com>
Siska C, Bowler R and Kechris K. The Discordant Method: A Novel Approach for Differential Correlation (2015), Bioinformatics. 32 (5): 690-696. Lai Y, Zhang F, Nayak TK, Modarres R, Lee NH and McCaffrey TA. Concordant integrative gene set enrichment analysis of multiple large-scale two-sample expression data sets. (2014) BMC Genomics 15, S6. Lai Y, Adam B-l, Podolsky R, She J-X. A mixture model approach to the tests of concordance and discordancd between two large-scale experiments with two sample groups. (2007) Bioinformatics 23, 1243-1250.
## load Data data(TCGA_GBM_miRNA_microarray) # loads matrix called TCGA_GBM_miRNA_microarray data(TCGA_GBM_transcript_microarray) # loads matrix called TCGA_GBM_transcript_microarray print(colnames(TCGA_GBM_transcript_microarray)) # look at groups groups <- c(rep(1,10), rep(2,20)) ## DC analysis on only transcripts pairs vectors <- createVectors(TCGA_GBM_transcript_microarray, groups = groups) result <- discordantRun(vectors$v1, vectors$v2, TCGA_GBM_transcript_microarray) ## DC analysis on miRNA-transcript pairs vectors <- createVectors(TCGA_GBM_transcript_microarray, TCGA_GBM_miRNA_microarray, groups = groups, cor.method = c("pearson")) result <- discordantRun(vectors$v1, vectors$v2, TCGA_GBM_transcript_microarray, TCGA_GBM_miRNA_microarray)