DESeq2 - Struggling to Define Desired Contrast

Question

[robert.flynn.21]

Hi all,

I am currently analysing some RNA seq data in which I have cases and controls which have each been treated with either glucose or galactose.

I would like to run two tests.

1. Compare glucose and galactose treatment within controls.
2. Compare glucose and galactose treatment within cases.

However, I have been struggling to define both of these contrasts.

For context, my metadata contains the following variables; condition (case/control), treatment (glucose/galactose), condition_treatment. (a combination of the condition and treatment variables which has 4 levels)

Below is the code in which I attempt to compare treatment within controls:

dds_glu_gal <- DESeqDataSetFromMatrix(counts, meta, design = ~condition_treatment)

dds_glu_gal <- DESeq(dds_glu_gal)

#Define contrast
contrast_condition <- c("condition_treatment", "control_glucose", "control_galactose")

#Extract results
res_table_control_gg <- results(dds_glu_gal, contrast = contrast_condition, alpha = 0.05)


#Apply LFC shrinkage 
resultsNames(dds_glu_gal)
[1] "Intercept"                                              
[2] "condition_treatment_case_glucose_vs_case_galactose"     
[3] "condition_treatment_control_galactose_vs_case_galactose"
[4] "condition_treatment_control_glucose_vs_case_galactose"

As you can see, there is no contrast which compares glucose and galactose treatment within controls. At first I though this contrast may be contained within the intercept term. However, the reference level for the condition_treatment variable is case_glucose. If have not misunderstood, then the intercept term should be defined by the factor reference level?

Would you have any idea what I can do? I'm fairly sure I have missed something basic!

Many thanks,
Rob

Answer 1

[Ram]

What does typing in just res_table_control_gg on the R command line print? You're comparing resultNames(dds_glu_gal) to your requirement. The result names shown have nothing to do with the lines before that involving the contrast_condition variable, the result names are from the DESeq(dds_glu_gal) call.

[robert.flynn.21]

Ah yes of course. It prints out

log2 fold change (MLE): condition_treatment control_glucose vs control_galactose 
Wald test p-value: condition_treatment control_glucose vs control_galactose 

DataFrame with 26485 rows and 6 columns

and then a bit of the dataframe.

Silly me, it looks like the comparison I wanted was already defined.

Thankyou for your help.

[Ram]

Please accept my answer to provide closure to the post.