RNAseq Normalization for eQTL Mapping?

Question

[QVINTVS_FABIVS_MAXIMVS]

How do you normalize RNAseq data for eQTL mapping?

I have 200 samples with bulk RNAseq data, from that I generated matrix of TPM values for each gene. I also have a matrix of expected alignment counts and a matrix of read counts.

I suppose I should use TPM for the eQTL mapping. If so, how do I normalize the values? Do I perform a quantile normalization or some other method? Should I even normalize the TPM matrix?

Additionally, what TPM cutoff should I apply and should I exclude genes according to X% of samples with TPM expression over Y_TPM_CUTOFF?

Thanks!

[Kevin Blighe]

It will likely depend on the eQTL program that you're using and the type of data distribution that it expects. Which one are you aiming to use?

[rodd]

Hi Kevin! [Sorry for jumping in on this question... :P ] I would like to perform an eQTL mapping analysis using QTLtools or fastQTL, but I am not sure how I should normalize or transform the count/expression data. [I noticed that if I use just regular DESeq2-normalized values, or variance stabilized transformed values, the results are pretty similar. But which one is best to use (and why)?]