Find DEG between sample with replicates and sample without replicates

Question

[l0o0]

Hi all. I just try to find differentially expressed genes between sample with replicates and sample without replicates. I know this sound weird, but I think it's better than nothing. Does anyone have encountered this problem and give some advices?

[bioinforesearchquestions]

Hi friends,

I have 15 pairs of samples, each pair has one cancer and one normal sample. Will this have a bigger impact on my analysis and interpretation?

Answer 1

[Farbod]

Dear linxzh1989, Hi.

Do you intend to compare DEG between 2 conditions = case and control (e.g 3 biological replicates of cancer samples and just one healthy sample) or just comparing 3 cancer replication samples Vs. one cancer sample ?

In Trinity DEG pipeline it has used a "--dispersion" factor for "No Biological Replicates" condition.

[ATpoint]

While technically possible, you have 0% guarantee that the factor you enter is correct. Technical confounders might inflate that value, the only reliable source of dispersion estimates are replicates.

Answer 2

[Carlo Yague]

In the DESeq manual, they suggest that you can do it using only the condition with replicates to estimate the dispersion. However this assume that the unreplicate condition doesn't have larger variation than the replicated one. In any case, I wouldn't draw strong conclusion from such an analysis but, yes, its better than nothing.

See section 3.2 of the DESeq manual.