Hi,

I am trying to analyze what genes are affected by the interaction factor in my model.

When I do 2 by 2 contrasts using Wald, the volcano plots look ok (V shape). When I want to plot the genes that were significantly affected by the interaction -reducing it in the LRT- (adjusted p-value this time), then I get noise in the central area of the volcano plot, getting kind of an inverted V shape. I am using -log10(adjp) in LRT, and -log10(pval) in Wald.

Why is that happening? How is that genes with no much fold change still get high adjusted p-values in LRT? If I put it in words, genes that have no significant fold change when compared with the reference condition (log2FC close to 0), still get a significant impact from the interaction factor.. (or whatever factor I test with LRT).

Am I plotting this wrongly?

Thanks,

Laia

See the information in vignette and ?results about LRT. The coefficient/LFC is just one of many that are represented in the LRT.