Correlating miRNA count from DeSEQ2 with Transcript FPKM from cuffdiff

Question

[arisarkar88]

Hi, I am trying to do a correlation study with miRNAs and its target transcript. The problem lies in the fact that the normalized counts for miRNAs are obtained from DeSeq2 whereas the normalized transcript counts (i.e. FPKM) are from Cuffdiff analysis. I do not have the raw data and as such cannot redo the analysis for transcript using DeSeq2 as that would have been the best way. Is there a way to bring them to the same scale of normalization and then go for correlation analysis? I even tried HT-Seq but I am not sure whether that will serve the purpose. Any suggestion please.

Answer 1

[Kevin Blighe]

Your best option may be this:

• transform the DESeq2-normalsied counts to variance-stabilised counts, and then Z-transform with t(scale(t(x)))
• Z-transform the FPKM counts via zFPKM R package

Then, at least, they will be on the same distribution.

You could also consider a PCA-based transformation and then correlate loadings / eigenvectors.

Kevin