Hi, I am conducting a DESeq2 differential expression analysis for which I have 200 tumor samples and 170 of their paired normal tissues.

We intend to take them all (200) into the same article and I am wondering how can I analyze them together in a statistically acceptable way.

So far the idea is to do the 170 tumor vs 170 normal paired test (170 patients) and then a unpaired 30 tumor vs 30 normal (60 patients) test. For the unpaired analysis I would be using 30 samples taken randomly from the pool of 170, or those that are grouping near in a heatmap...does this makes sense?

Then, I would be comparing the log2foldchanges of the 200 cases together if possible in order to find subgroups.

Would this be acceptable?

Thank you for this fantastic tool.

David.

Hi David,

This has been asked before but I’m sure it’s very difficult to find. DESeq2 can only make use of the paired samples, while limma and its duplicateCorrelation function can make use of unpaired samples as well by modeling the pair correlation.