Comparing two differential expression experiments

Question

[user31888]

I ran independently 2 differential expression experiments using DESeq: . CONDITION_1 vs REFERENCE . CONDITION_2 vs REFERENCE

(REFERENCE is the same for the 2 CONDITIONS)

I got 2 tables like below showing the differentially expressed genes between my sample and the REF (1 table per CONDITION):

id              baseMean baseMeanA baseMeanB foldChange   log2FoldChange    pval
ENSG00000000005 3.5      1         6         6            2.584962501       0.445151971
ENSG00000000419 14.5    15        14         0.933333333  -0.099535674      1

Within each condition:

(1) Does a positive log2FoldChange means that the gene in the sample is over-expressed compared to the REF?

(and a negative log2FoldChange for down-expression?)

Between 2 conditions:

(2) To know which gene shows the largest difference in expression between CONDITION_1 and CONDITION_2, would it be correct to simply make the difference of the 2 respective log2FoldChange?

(i.e. log2FoldChange_CONDITION_1 - log2FoldChange_CONDITION_2)

Or could we do a ratio of the 2 p-values instead?

[shunyip]

(1) I usually just randomly pick a gene from the matrix and see for myself which condition has the higher expression. Then I'd know what the log2FoldChange means.

(2) To compare CONDITION_1 and CONDITION_2, isn't it easier to just perform another DE experiment?

[user31888]

(1) Don't get it, sorry. Not sure it would be helpful to randomly pick a gene. (2) Since in my case I have 3 samples: REFERENCE, sample-CONDITION_1, sample_CONDITION_2, DESeq does not allow me to compare more than 2 conditions (I have to assign the sample group before calculating the log2FoldChange).

[shunyip]

(1) Oh, what I meant was to first, randomly pick a gene that has a reasonable level of expression. Then, look at its expression in reference and condition 1. If you see that its expression went up/down from reference to condition 1 and if you see positive log2 fold change, then you'd know what a positive/negative log2 fold change mean. To make sure, you can choose several genes with higher fold change difference too, this isn't time consuming.

(2) I mean performing a DE experiment with 2 conditions: CONDITION_1 and CONDITION_2.