Hello,

I'm new to bio informatics and you love a little help to know if I'm doing the right thing. I'm using tcga data from breast cancer to perform some correlation between two gene signature. Here's how I proceeded :

I downloaded the data using GDC. Taken the tpm counts and log transformed them. then I used gsva score to score both my signature on all the samples. perform a pearson correlation.

My question is : Is there any pre processing I have missed. In the litterature some correlation are so strong that I'm almost sure they filter some samples based on sample purity or immune cells infiltrate. Of course one of my signature is on a type of immune cell so it strongly depends on sample purity.

One other analysis I have performed is separate between signature 1 high and low and see the scoring on the second signature.

But again I'm not quite sure if the pre processing is fair and square.