This is a test version of Biostars. For the public version, visit https://www.biostars.org.
Spearman correlation rna-seq gene to blood values

Hi everyone,

I have a very basic question. I try to correlate rna-seq gene expression levels to values from laboratory blood analysis. To do so, I would use spearman correlation. Now I worry about the normalization. Currently I am using DESeq2 with vst to normalize gene expression. Is that a correct approach? Is a normalization step needed at all, since spearman is a non-parametric rank test or could I also use raw or simple log2 normalized counts?

Thank you very much.

Best

Chris

Edit/Added: I just tried different normalisation methods. The bluish curve was correlated to a blood parameter that was expected to be completely independent. The redish curve is expected to correlate with a high number of genes. Even with that knowledge, I don't know what method represents the data in this case. Critical genes seem to be captured by all approaches, on the first sight. Density Plots of rho values for different normalisation methods

spearman rna-seq
• 1,495 views • link
updated by ATpoint ATpoint • written by bas2020 • bas2020

1 answer

I would use a method that does not alter the ranking of the original count data. In standard RNA-seq you sequence the full transcripts but if you use TPM you divide out gene length. That means in contrast to DESeq2 and edgeR this method alters the ranks as compared to the original counts. Just use DESeq2 normalized counts, that should be fine.

• 0 views • link
written by ATpoint ATpoint

Log in to answer this question.

More posts like this