Dear all,

I used Deseq2 to analyse differential gene expression: I have 2 factors:

- phenotype (2 levels: slow and fast)

- timepoint (3 levels: RT, 15, 90)

Overall, I want to know which genes are differentially expressed between the slow and the fast phenotype at each of the three timepoints. After going through the vignette, I decided to stick to "Example 2" at the results help page. That worked out well but I am not sure if that was the correct design for my purpose and also if I interpret the results output in the correct way.

I am using R version 3.3.0 (2016-05-03) and DESeq2_1.12.2

dds <- DESeqDataSetFromMatrix(countData= countData,
                              colData= sampleData,
                              design= ~phenotype+ timepoint + phenotype:timepoint) 
colData(dds)$timepoint <-factor(colData(dds)$timepoint, levels=c("RT", "15", "90")) 
colData(dds)$phenotype <-factor(colData(dds)$phenotype, levels=c("slow", "fast"))
dds1 <- DESeq(dds)

Now the results:

resultsNames(dds1)
[1] "Intercept"                 "phenotype_fast_vs_slow"    "timepoint_15_vs_RT"        "timepoint_90_vs_RT"       
[5] "phenotypefast.timepoint15" "phenotypefast.timepoint90"

Question1:

results(dds1, name = "timepoint_15_vs_RT") This should tell me how many genes are "up- or down-regulated" at timepoint 15 vs RT in the control phenotype ("slow") ONLY, right? (Not across all samples/both phenotypes...?)

Question 2:

results(dds1, contrast = list(c("timepoint_15_vs_RT" , "phenotypefast.timepoint15"))) This is the main effect + interaction term and should tell me how many genes are up or down regulated at timepoint 15 vs RT in the second phenotype ("fast"), right?

But what does up- and down- exactly mean in that case? Up-regulated compared to timepoint 15 in the control phenotype ("slow") OR up-regulated compared to timepoint RT in this phenotype ("fast")...?

Every comment/insight on that would be helpful for me! Thanks to everyone in advance!

Best,

Annabella 

Hello,

I am not an expert but I have done multiple factor before although simpler than yours which is comparison for 3 phenotypes.

Answering your question, I think fast vs slow is all data in fast category vs all data in slow category. There is no time separation. RT vs 15, RT vs 90, 15 vs  90 also same. All data in one time category vs all data in other time category. No fast/slow separation.

As for your goal, I think what you want to do is still comparison of 2 phenotypes, fast vs slow but for different time. In that case, the easier way is to compare it by grouping the data for each time, RT, 15, and 90. Maybe you can have 3 matrices representing these groups, for example TimeRTMatrix, Time15Matrix, and Time90Matrix. Each matrix consists of samples fast and slow samples. Then you just need to process each matrix one by one. You will have 3 DESeq results and each results will give you comparison of fast vs slow for each time point. In my opinion this is the simplest way to do it although take longer time because you must process 3 times.

hi Annabella,

You've got it right:

"Question1: results(dds1, name = "timepoint_15_vs_RT") This should tell me how many genes are "up- or down-regulated" at timepoint 15 vs RT in the control phenotype ("slow") ONLY, right? (Not across all samples/both phenotypes...?)"

Yes, this is the 15 vs RT comparison for pheno=slow only.

"Question 2: results(dds1, contrast = list(c("timepoint_15_vs_RT" , "phenotypefast.timepoint15"))) This is the main effect + interaction term and should tell me how many genes are up or down regulated at timepoint 15 vs RT in the second phenotype ("fast"), right?"

Yes, this is the 15 vs RT comparison for pheno=fast only.

"But what does up- and down- exactly mean in that case? Up-regulated compared to timepoint 15 in the control phenotype ("slow") OR up-regulated compared to timepoint RT in this phenotype ("fast")...?"

The latter. Up-regulated at timepoint 15 compared to RT only for pheno=fast.

Dear Michael and bharata,

thank you a lot for your answers! It is clear for me now!