Matched samples in Deseq2

Question

[foejvs546]

Hi everyone,

I am doing an rna-seq project where I am looking at gene expression in cancer.

I have two sample groups: healthy and cancer. The samples come from biopsy tissues - tumour and healthy tissue adjacent to the tumour.

Since I am only interested in cancer vs healthy, does it matter that each sample is matched by patient? Should I account for this in Deseq2 somehow?

Thank you for your patience if this is a naive question!

Answer 1

[Ram]

As long as you have replicates, you can contrast tumor vs normal but remember that the results will be a comparison of averages and you cannot make any individual observations from them. You need matched normal for somatic variant/CN/SV calling but RNAseq won't suffer much from a bulk/pooled approach.

[foejvs546]

Hi Ram, thank you so much for your input.

I am only interested in average 'healthy vs cancer' changes. Is there a way to use the matching information to help?

Sorry if this sounds silly, someone said to me there's a way to improve the power by accounting for the matchedness in deseq. I don't know if that is at all accurate.

[Ram]

I think I read your question wrong, so my answer is not relevant. Wait for someone else's take. In my opinion, the only way to account for matchedness in DESeq2 is to add a "patient" covariate but I'm not sure how that will help as each patient will only have one cancer/healthy sample, ensuring a complete lack of replicates when you use a ~ diseaseState + patient formula. Wait for someone else's take, I'm beyond my depth here.

[foejvs546]

Hi Ram, Ok thank you very much.