DESeq2 - finding influence of multiple parameters

[MagdalenaZ]

I have a dataset which is multivariate and balanced, i.e.

"~Cells*Treatment"

Cells: A or B
Treatment: yes or no

Is there any way I can - for each gene - calculate how much of its expression is determined by "Cell" and by "Treatment"?

For instance; expression of gene1 is determined 30% by Cell-type and 70% by treatment?

Cheers,
Magdalena

[Michael Love]

You could compare the likelihood ratio test of ~cells + treatment + cells:treatment to ~cells. The likelihood ratio test statistic gives a sense of the extra explanatory power of adding information regarding treatment to the model. You could do the reverse for cells. This doesn't give you percent but you can visually compare the two statistics in a plot.

[MagdalenaZ]

Great idea, I'll give that a try.
Thank for your help Michael!