dseq2 results synergy contrast
1
0
Entering edit mode
steven wink ▴ 90
@steven-wink-5440
Last seen 2.4 years ago

Hi!

I am having some difficulty defining the correct contrast to test for synergy of two treatments A & B vs AB using the deseq2 results function. Treatment has 4 levels A, B, combined AB and reference level control.

When I set up contrasts e.g. ( AB - control ) - [(A - control) + (B - control)] I end up with AB - A - B + control! How can I get rid of the double counted control here? What am I missing? My design is ~ donor + condition + treatment + treatment:condition

deseq2 contrast synergy • 307 views
1
Entering edit mode
@mikelove
Last seen 7 hours ago
United States

Can you provide your colData, and more description of what you are looking to test for? You haven't described condition yet.

0
Entering edit mode

Thanks for helping Michael, Condition has two levels. I want to test if the combined AB treatment is different from the additive A and B treatment (AB - A - B) within either conditions. First in the reference condition, later adding the interaction term to test within the hypoxia condition.

sample(rowname)         condition   treatment   donor
HSF1_hypoxia_A          hypoxia     A           HSF1
HSF1_hypoxia_AB         hypoxia     AB          HSF1
HSF1_hypoxia_B          hypoxia     B           HSF1
HSF1_hypoxia_control    hypoxia     control     HSF1
HSF1_normoxia_A         normoxia    A           HSF1
HSF1_normoxia_AB        normoxia    AB          HSF1
HSF1_normoxia_B         normoxia    B           HSF1
HSF1_normoxia_control   normoxia    control     HSF1
HSF2_hypoxia_A          hypoxia     A           HSF2
HSF2_hypoxia_AB         hypoxia     AB          HSF2
HSF2_hypoxia_B          hypoxia     B           HSF2
HSF2_hypoxia_control    hypoxia     control     HSF2
HSF2_normoxia_A         normoxia    A           HSF2
HSF2_normoxia_B         normoxia    B           HSF2
HSF2_normoxia_control   normoxia    control     HSF2
HSF3_hypoxia_A          hypoxia     A           HSF3
HSF3_hypoxia_AB         hypoxia     AB          HSF3
HSF3_hypoxia_B          hypoxia     B           HSF3
HSF3_hypoxia_control    hypoxia     control     HSF3
HSF3_normoxia_A         normoxia    A           HSF3
HSF3_normoxia_AB       normoxia     AB          HSF3
HSF3_normoxia_B         normoxia    B           HSF3
HSF3_normoxia_control   normoxia    control     HSF3
HSF4_hypoxia_A          hypoxia     A           HSF4
HSF4_hypoxia_AB         hypoxia     AB          HSF4
HSF4_hypoxia_B          hypoxia     B           HSF4
HSF4_hypoxia_control    hypoxia     control     HSF4
HSF4_normoxia_A         normoxia    A           HSF4
HSF4_normoxia_AB        normoxia    AB          HSF4
HSF4_normoxia_B         normoxia    B           HSF4
HSF4_normoxia_control   normoxia    control     HSF4
HSF5_hypoxia_A          hypoxia     A           HSF5
HSF5_hypoxia_AB         hypoxia     AB          HSF5
HSF5_hypoxia_B         hypoxia      B           HSF5
HSF5_hypoxia_control    hypoxia     control     HSF5
HSF5_normoxia_A         normoxia    A           HSF5
HSF5_normoxia_AB         normoxia   AB          HSF5
HSF5_normoxia_B         normoxia    B           HSF5
HSF5_normoxia_control   normoxia    control     HSF5

0
Entering edit mode

Easiest is to code this as colData with the columns: condition, A (ctrl/trt), and B (ctrl/trt).

Then your design is ~condition + A + B + condition:A + condition:B + A:B + condition:A:B

0
Entering edit mode

Not sure I fully understand, could you explain a bit more in detail? Did you note that I have 4 levels in my treatment column? (A, B, AB and control) As I read it now I would create extra dummy columns in colData consisting of ones and zeros, with column A having entries as ones for when treatment is control or treatment A, and column B having entries as ones for when treatment is control or B. On paper I would not be able to set up the correct contrast? I'd get this I think:
(A + control - B - AB) - (B + control - A - AB) so I still do not end up with A + B - AB.

0
Entering edit mode

The design I suggested will allow you to assess whether there is an interaction with A and B, and whether this is same or different across condition. As it’s a complex design, if you’re not certain about the terms and how to connect them to interpretations, I’d recommend you to collaborate with a local statistician.

Traffic: 393 users visited in the last hour
FAQ
API
Stats

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.