Entering edit mode
cryptic
•
0
@cryptic-20547
Last seen 4.7 years ago
I have searched on both bioconductor as well as biostars but haven't found very relevant results. My experimental conditions setup is as follows:
sample strain condition
samp01 WildType time0
samp02 WildType time0
samp03 WildType time0
samp04 WildType time1
samp05 WildType time1
samp06 WildType time1
samp07 WildType time2
samp08 WildType time2
samp09 WildType time2
samp10 mutation_1 time0
samp11 mutation_1 time0
samp12 mutation_1 time0
samp13 mutation_1 time1
samp14 mutation_1 time1
samp15 mutation_1 time1
samp16 mutation_1 time2
samp17 mutation_1 time2
samp18 mutation_1 time2
samp19 mutation_2 time0
samp20 mutation_2 time0
samp21 mutation_2 time0
samp22 mutation_2 time1
samp23 mutation_2 time1
samp24 mutation_2 time1
samp25 mutation_2 time2
samp26 mutation_2 time2
samp27 mutation_2 time2
Essentially, WildType is my CONTROL, and mutation1 and mutation2 are experimental conditions. Each of these three strains were sampled at 3 time intervals each. Given this data, can I setup the analysis as follows?
data_all <- DESeqDataSetFromMatrix(countData=mycounts, colData=mypheno, design = ~ strain + condition + strain:condition)
Michael: Thanks for your swift response. I will take a look at the workflow on bioconductor. In the meantime, I went ahead and ran the model as above. As per
resultsNames(data_all_deseq), I have 9 comparisons. But if I look at the results table, I see only onepadjcolumn. Thus, it looks like the p-value is being calculated over all conditions - which seems wrong.So I probably need to use the
contrastoption to explicitly compare pairs of conditions. For example:WildType_time0vsmutation_1_time0I’d recommend working with a statistician if it’s your first time working with linear models. Or you can take at look at the workflow and vignette which are fairly extensive. I unfortunately don’t have too much time for statistical questions and have to reserve my time for software related questions.