DESeq2 multifactorial design

[sisterdot]

I hope someone can help me out in defining the right design formulas and contrasts to use with DESeq2 for a (subjectively) rather complex experimental design.

The two infection types infA and infB were considered to be equivalent by the experimental scientists.

The experimental design was as follows:

	genotype	sex	inf_type	inf
m_KO_iA	KO	m	infA	inf
m_KO_1	KO	m	NOinf	NOinf
m_KO_iB	KO	m	infB	inf
m_KO_2	KO	m	NOinf	NOinf
m_WT_iA	WT	m	infA	inf
m_WT_1	WT	m	NOinf	NOinf
m_WT_iB	WT	m	infB	inf
m_WT_2	WT	m	NOinf	NOinf
f_KO_A	KO	f	infA	inf
f_KO_1	KO	f	NOinf	NOinf
f_KO_iB	KO	f	infB	inf
f_KO_2	KO	f	NOinf	NOinf
f_WT_iA	WT	f	infA	inf
f_WT_1	WT	f	NOinf	NOinf
f_WT_iA	WT	f	infA	inf
f_WT_2	WT	f	NOinf	NOinf

And the questions i would like to pose with DESeq2 are:
- how do the KO cells differ in their response to any treatment compared to WT cells.
- how do the KO cells differ in their response to treatment A (infA) compared to WT cells.
- how do the KO cells differ in their response to treatment B (infB) compared to WT cells.

any suggestions on design formulas and contrasts to use are very much appreciated.
i am really sorry for not providing a suggestion myself, but i am afraid that might get embarrassing

THANKS

[Wallysb01]

You should read up on the interaction terms and changing the comparisons when you call the results function.

The design probably needs to look something like:

design ~ genotype + sex + inf_type + inf + sex:inf + sex:inf_type + genotype:inf + genotype:inf_type

By default the results function will test the last term in your design function, so you can change that or just use the comparisons option with results.

[sisterdot]

Thanks for your reply...

i am going for
~ genotype + sex + inf + sex:inf + genotype:inf
for now...

gave a larger set of significant genes than not accounting for sex
~ genotype + inf + genotype:inf

is it possible to check in advance which main effects and interactions are best to include ?