Question

Understandig listValues argument of DESEQ in multiple group comparisons

0

Entering edit mode

User000 • 0

@ea03770f

Last seen 2.1 years ago

Italy

Dear all,

I would like to understand the listValues argument of DESEQ2 when comparing multiple groups. For i.e. 1 vs 2 is like this,

res <- results(dds, contrast=list("conditionAB", c("conditionAA", "conditionBB"), listValues=c(1, -1/2))

What about 4 vs 2? or 2 vs 3?

res <- results(dds, contrast=list(c("conditionA","conditionB","conditionC","conditionD"), c("conditionAA", "conditionBB")), listValues=c(4, -1/2))

res <- results(dds, contrast=list(c("conditionA","conditionB"), c("conditionAA", "conditionBB","conditionCC")), listValues=c(2, -1/3))

Is this right?

DESeq2 • 1.9k views

ADD COMMENT • link updated 20 months ago by Michael Love 43k • written 2.8 years ago by User000 • 0

score 1 · Answer 1 · 2022-02-18

1

Entering edit mode

Michael Love 43k

@mikelove

Last seen 4 days ago

United States

It's important to know what the coefficients mean. If you have a multi-group design and you have excluded the intercept, then you are just averaging the cell means. So 2 vs 3 should be 1/2 and -1/3. But the correct answer here depends on the design.

ADD COMMENT • link 2.8 years ago Michael Love 43k

0

Entering edit mode

I think I am not understanding at all this part. Could you please explain better, otherwise forward me to some link to understand better. Thnaks

ADD REPLY • link 2.8 years ago User000 • 0

0

Entering edit mode

Unfortunately I don’t have a lot of extra time so I have to limit myself to software related questions. For statistical design and interpretation of results I’d recommend collaborating with a local statistician or someone familiar with linear models in R.

ADD REPLY • link 2.8 years ago Michael Love 43k

0

Entering edit mode

Thank for your time and reply. If I had a local statistician I wouldn't be asking this question in community. Unfortunately, in the manual I do not find a lot of information or examples, which would help to understand particular cases. Only classical treated vs untreated. I have treated data of several patients (no replicates) which I want to compare together against untreated donor patients. Since these are not replicates but data from different patients at different level of disease, I guess taking the average is a good way? This is what I am doing, but I am not sure this is the rigth way to compare conditions coming from different patients. Any suggestion is much appreciated. Thanks.

mm = model.matrix(~0+condition, sampledata)
dds <- DESeqDataSetFromTximport(txi.g, colData=sampledata, design=mm)
dds <- DESeq(dds)
res <- DESeq2::results(dds, 
               contrast = list(c("condition_treated_patient1","condition_treated_patient2","condition_treated_patient3","condition_treated_patient4"),c("condition_control_patient1","condition_control_patient2")), 
               listValues = c(1/4,-1/2)
)

ADD REPLY • link 2.8 years ago User000 • 0

0

Entering edit mode

Despite not having coordinated someone you could plan your statistical analysis with, I still can't offer statistical consulting on the support site. I have to restrict my time to software related questions.

ADD REPLY • link 2.8 years ago Michael Love 43k

score 0 · Answer 2 · 2022-02-18

0

Entering edit mode

swbarnes2 ★ 1.4k

@swbarnes2-14086

Last seen 2 days ago

San Diego

You can do just about what you've written much simpler by making a new column of ColData with just "Condition_ABCD" and "Condition_AABB", for example. It's not exactly the same as making lists in the contrast like you did, but it sure is a lot easier to read and understand, and harder to mess up.

ADD COMMENT • link 2.8 years ago swbarnes2 ★ 1.4k

score 0 · Answer 3 · 2023-03-06

Hi Michael Love,

Do you have some example on how to run listValues argument with batch correction? I want to do 1vsALL calculation for my data (18 conditions) and but the data also have batch effect (5 batches). based on previous post, I plan to run my DE like this:

rds <- results(dds, contrast = list(c(group1),c(othergroups)), listValues = c(1,-1/17))

However, I keep having problem because group1 is missing.

this is the result if I run resultsNames(dds):

"batch1", "batch2", "batch3", "batch4", "batch5", "group2", "group3", "group4", "group5", "group6", "group7", "group8", "group9", "group10", "group11", "group12", "group13", "group14", "group15", "group16", "group17", "group18"

This is how I run my code:

dds <- DESeqDataSetFromMatrix(countData = input, colData = group, design= ~ 0+batch+group)

Any suggestion on how should I put on my design?