Hi there
I have a large RNAseq dataset where the expression of about 20,000 genes is determined for various different B-cell populations. I have performed WGCNA and have identified consensus modules using all conditions.
My next task is to see how these modules correlate with 'traits' (module-trait relationship) and to display these correlations in a heatmap, to identify which modules are associated with the various B-cell types.
Now comes my problem. can I use the gene expression values for the different B-cell categories as trait data? I have tried using the following code:
moduleTraitCor = cor(MEs, datTraits, use = "p"); moduleTraitPvalue = corPvalueStudent(moduleTraitCor, nSamples)
labeledHeatmap(Matrix = moduleTraitCor, xLabels = rownames(datTraits), yLabels = names(MEs), ySymbols = names(MEs), colorLabels = FALSE, colors = blueWhiteRed(50), textMatrix = textMatrix, setStdMargins = FALSE, cex.text = 0.4, zlim = c(-1,1), main = paste("Module-trait relationships"), cex.lab.y = 0.5)
the problem is that I have generated a heatmap displaying correlations for module eigengene values with expression level for ALL genes - I want to see the correlation between module eigengene values for the different B-cell groups
How can I do this?
My RNAseq expression matrix provides gene expression values for all genes for different B-cell subtypes - so germinal centre B-cells, Marginal zone B-cells, Follicular B-cells etc. matrix = rows (genes) x columns (B-cell group)
You can correlate the module eigengenes to anything, whether continuous or categorical. If you want to correlate to a group of genes at the same time, you will have to come up with a summary metric for those genes.
Dear Kevin and joetaylor268,
I have very similar question to the WGCNA-related matter discussed above. Thus, only for the clarification:
As a trait , can I directly use my gene of interest's TPM value ? i mean , can I just copy the TPM values for my gene and paste it in the trait file to see co-expressed or correlated genes? In my case it is only one gene. @joetaylor268 : Did your analyses return biologically relevant results with this setup? Many thanks in advance for your reply.
Yes, you can use anything, but this same gene was used to construct the modules, no? You may just want to be careful about how you interpret the result. You could also 'factorise' your gene into, e.g., tertiles or quartiles, and then use that in the correlation. Then you would have, e.g., GENE_X-hi correlates with Pink Module, but GENE_X-lo correlates with Black Module.
Thank you for your quick reply. Yes, the same gene was in the input files, amongst 10000 other expressed genes. In short: basically GENE_X is always upregulated in condition#B compared to condition#A. Thus, I crated a trait file, where I have a column for GENE_X and the column contains the exact TPM values for each sample (extracted from the gene expression matrix-input file). I will follow your advice and will repeat the analyses where I will have two columns for the GENE_X: HI and LO.