comparing DEGs with same treatment but different control batch

Question

[firatuyulur]

Hi,

Its sound quite trivial but I couldnt find the way somehow. I have an RNASeq with 3 replicates of treatment and 9 controls. As I noticed something strange with controls looking at their raw counts by eye, I normalised and clustered them. They clustered in 3 different groups. Even this tells me a difference between them anyways, I wanted to see whether their deseq2 output would look any similar. I picked 2 of the 3 clusters and ran deseq2. So the structure is as TreatmentRep1 TreatmentRep2 TreatmentRep3 Control1 Control2 Control3 and TreatmentRep1 TreatmentRep2 TreatmentRep3 Control4 Control5 Control6 . Now what I want to see is correlation between the two deseq2 outputs keeping the geneNames as my constant point. Probably the best way to compare them would be on log2fc values. Accordingly what I have is a dataframe like this

geneNames    L2FC_Comp1    L2FC_Comp2
GeneA            x1            y1
GeneB            x2            y2
GeneC            x3            y3
GeneD            x4            y4
...
...

Many statistical correlation methods use the whole population and their mean or another parameter to compare but in this case I also care about the geneNames that the comparison should consider each geneName for correlation.

Answer 1

[swbarnes2]

You can convert the results to data frames, then you can merge by gene names. But I'm not sure that's how you should do it. I think you should add a new column for batch, and use that in your design. And if your treatments are in a different batch than your controls, you have a real problem. Even if you find gene changes that differ between your treatments and all three sets of control,s you will have no idea if it's because of the treatment, or because of the batch effect.