Hello there,

I guess I would like some clarifications on how DESeq2 handles multiple testing in terms of contrasts of interest along with contrasts of other variables. For example, please see the resultsNames() output from a paired design below:

resultsNames(deseq)  [1] "Intercept"                "Subject2vs1"      [3] "Subject3vs1"     "Subject4vs1"      [5] "Subject5vs1"     "Subject6vs1"      [7] "Subject7vs1"     "Subject8vs1"      [9] "Subject9vs1"     "Subject10vs1"     [11] "ConditionBvsA" "ConditionCvsA"  [13] "ConditionDvsA"  "ConditionEvs_A"

In this example, the contrasts of interest is denoted by the coef [11] to [14], the multiple testing correction only takes into account the number of genes in each contrast or the number of correlated contrasts (e.g. [11]-[14]) are also being corrected? It looks to me that the former is the case; if so, is there a way that the latter can be achieved in DESeq2 to account for the 4 contrasts [11]-[14] that may be correlated in the example above?

Equivalent to this, if lfcShrink() function is used to summarise the results, how the s-value probability may be considered in relation to multiple testing correction?

Thanks for your help. Guan

The adjusted p-value column only correct across genes. (Same for s-value— these are generated one at a time.)