Hi all,
I have done a time series analysis using the steps outlined below and I got differentially expressed genes showing a condition-specific effect (according to the example at http://www.bioconductor.org/help/workflows/rnaseqGene/#time)
> sampleTable<-data.frame(sampleName=sampleFiles, fileName=sampleFiles, condition=sampleCondition, time=time)
> sampleTable$condition <- relevel(sampleTable$condition, "Valor")
> sampleTable$time <- factor(sampleTable$time, levels=c("0","6","12","24","72"))
> ddsHTSeq<-DESeqDataSetFromHTSeqCount(sampleTable=sampleTable, directory=directory, design=~condition + time + condition:time)
> ddsTS2 <- DESeq(ddsHTSeq, test="LRT", reduced = ~ condition:time)
> resultsNames(ddsTS1)
[1] "Intercept" "condition_BP1_vs_Valor" "time_6_vs_0"
[4] "time_12_vs_0" "time_24_vs_0" "time_72_vs_0"
[7] "conditionBP1.time6" "conditionBP1.time12" "conditionBP1.time24"
[10] "conditionBP1.time72"
> summary(resTS1)
out of 1840 with nonzero total read count
adjusted p-value < 0.1
LFC > 0 (up) : 138, 7.5%
LFC < 0 (down) : 113, 6.1%
outliers [1] : 4, 0.22%
low counts [2] : 0, 0%
(mean count < 0.2)
But now I am interested in picking the differential genes between the 2 conditions at each time-point, how do I get these? which contrast should I use specifically and which elements from my resultsNames(ddsTS1) output. I checked ?results but still not sure whether to use the condition effect or the interaction effect?
Many thanks,
Stan
So, I found the answer for my question in one of the Deseq2 threads at the bioconductor support forum, a nice explanation is given and the thread is found here: https://support.bioconductor.org/p/65676/#66860 in case it might be of use to someone else.
Hi Stan,
This is a bit late reply, but I am facing actually the same problem. So the answer is basically to separate your time-points and test each and every on certain condition? This seems to be tedious especially if you have a lot of time points!
Cheers, Deni
I guess you misunderstood what I tried to say, or I haven't explain myself well. Basically I want to test whether there is difference in my microbial community composition between different treatments in a time-series. So i have oil incubation over the period of 64 days. I have sampled at day 0, 3, 6, 9, 13, 16, 32 and 64. What I want to see is whether there is difference at each time-point between oil incubation and control. I was hoping to do it at once, but I could not find a solution for that. So my approach would be to "separate" each day and then do the testing control vs oil incubation basically on 8 different dataframes. Thanks for the response Deni
Log in to answer this question.
