Question

DESeq2 time-series data of multiple treatments with common time 0

0

Entering edit mode

Ting • 0

@ting-12258

Last seen 7.9 years ago

Hi all,

I'd like to use DESeq2 for treatment-specific differences over time.

Here is my colData with a common time 0 for Mock, treatment1 and treatment 2:

Control_a Control_b Control_c Mock_03hr_a Mock_03hr_b Mock_03hr_c Mock_06hr_a Mock_06hr_b Mock_06hr_c Mock_12hr_a Mock_12hr_b Mock_12hr_c treatment1_03hr_a treatment1_03hr_b treatment1_03hr_c treatment1_06hr_a treatment1_06hr_b treatment1_06hr_c treatment1_12hr_a treatment1_12hr_b treatment1_12hr_c treatment2_03hr_a treatment2_03hr_b treatment2_03hr_c treatment2_06hr_a treatment2_06hr_b treatment2_06hr_c treatment2_12hr_a treatment2_12hr_b treatment2_12hr_c operator">> as.data.frame(colData(cds)) times treatments 00 Mock 00 Mock
00 Mock
03 Mock
03 Mock
03 Mock
06 Mock
06 Mock
06 Mock
12 Mock
12 Mock
12 Mock
03 treatment1
03 treatment1
03 treatment1
06 treatment1
06 treatment1
06 treatment1
12 treatment1
12 treatment1
12 treatment1
03 treatment2
03 treatment2
03 treatment2
06 treatment2
06 treatment2
06 treatment2
12 treatment2
12 treatment2
12 treatment2

When I tried this:

cds <- DESeqDataSetFromMatrix(countData = data, colData = colData, design = ~ treatments+times+treatments:times)

I got the error message:

Error in checkFullRank(modelMatrix) : the model matrix is not full rank, so the model cannot be fit as specified. One or more variables or interaction terms in the design formula are linear combinations of the others and must be removed.

I've been searching for the reasons and solutions like this post: With DESeq2 "Not full rank" Error with design ~ line + time + condition, and went through its comments. However, since my data is unpaired, using:

colData$times.nested = factor(c(rep(1,3),rep(rep(2:5,each=3),3)))

cds <- DESeqDataSetFromMatrix(countData = data, colData = colData, design = ~ treatments + times + treatments:times.nested + treatments:times)

or

mm = model.matrix(~ treatments + times + treatments:times, colData(cds))

cds = DESeq(cds, full=mm, betaPrior=FALSE)

gave me the same error.

How do I capture differentially expressed genes of each treatment vs. Mock over time with common time 0 incorporated?

Best regards,

Ting

deseq2 multiple time points multiple treatments • 4.2k views

ADD COMMENT • link updated 7.9 years ago by Michael Love 43k • written 7.9 years ago by Ting • 0

score 1 · Answer 1 · 2017-05-15

1

Entering edit mode

Michael Love 43k

@mikelove

Last seen 4 days ago

United States

hi Ting,

There was a similar question posted recently, which didn't have any treatment samples at time 0. (There was also a second level of grouping in that question, but I didn't address this in my reply to the question.) Can you take a look at my answer and then see if you have further questions:

A: deseq2 multifactor timeseries help

For building the DESeqDataSet you can use a design of ~1, and then you can use matrices for the arguments 'full' (and 'reduced' if you want to perform a likelihood ratio test).

ADD COMMENT • link 7.9 years ago Michael Love 43k

0

Entering edit mode

Hi Michael,

With your solution, it works well now! Thank you so much for the prompt reply.

ADD REPLY • link 7.9 years ago Ting • 0

0

Entering edit mode

Hi Michael,

As I checked the comparisons of the analysis, it showed:

> resultsNames(cdsFull)
"Intercept" "times03" "times06"                
"times12" "times03.treatmentstreatment2" "times06.treatments treatment2"
"times12.treatments treatment2" "times01.treatmentsYeast" "times03.treatments treatment1"
"times06.treatments treatment1" "times12.treatmentstreatment1"

Does that shows there's no statistical comparisons between treatments as you said in A: deseq2 multifactor timeseries help ? Can genes with p.adj<0.05 in "times03.treatmentstreatment2" be interpreted as significant differences between treatment2 and Mock at times03 or just significant differences in treatment2 at times03 over time? If it's the latter case, is there other way to do the differential analysis between one of the treatments and Mock over time including common time 0? I once thought copy-paste the data of common time 0 to each treatment, thus the data set will be paired. But I guess this will change size factors?

Or sacrifice time 0 by removing the common time 0 and then just compare treatments over times03, times06, and times12.

ADD REPLY • link 7.9 years ago Ting • 0

0

Entering edit mode

times03.treatmentstreatment2 => treatment 2 vs mock at time 3

To compare treatment 2 vs treatment 1 at time 3, you would use the 'contrast' argument:

res <- results(dds, contrast=list("times03.treatmentstreatment2", 
                                  "times03.treatmentstreatment1")

ADD REPLY • link 7.9 years ago Michael Love 43k