[ashuchawla]

I have ten samples total, 2C10 = 45 combinations

BTW, what did you mean by "pairwise comparisons would take me a long time"? Can't be more than a few minutes calculation time. The issue is rather: what would the result tell you?

[Simon Anders]

45 comparisons still would only take an hour or so. However, my feeling is that you are doing something fundamentally wrong if you are comparing pairs of samples, rather than pairs of conditions or time points.

Maybe explain a bit more about the biology of your experiment.

[ashuchawla]

day 0
sample 1 - untreated

day 3
sample 2 - treated with "A"
sample 3 - treated with "B"
sample 4 - treated with "C"

day 6

sample 5 - treated with "A"
sample 6 - treated with "B"
sample 7 - treated with "C"

day 9

sample 8 - treated with "A"
sample 9 - treated with "B"
sample 10 - treated with "C"

They want to see DE across time as well as across treatments. I am just a bioinformatician. Some body else did the experiments and I have the rnaSeq data for it. This is all I have been told.

Ashu

45 comparisons still would only take an hour or so. However, my feeling is that you are doing something fundamentally wrong if you are comparing pairs of samples, rather than pairs of conditions or time points.

Maybe explain a bit more about the biology of your experiment.

[lucer105]

day 0
sample 1 - untreated

day 3
sample 2 - treated with "A"
sample 3 - treated with "B"
sample 4 - treated with "C"

day 6

sample 5 - treated with "A"
sample 6 - treated with "B"
sample 7 - treated with "C"

day 9

sample 8 - treated with "A"
sample 9 - treated with "B"
sample 10 - treated with "C"

They want to see DE across time as well as across treatments. I am just a bioinformatician. Some body else did the experiments and I have the rnaSeq data for it. This is all I have been told.

Ashu

I think you need a "untreated" control for every time point, or the time will be a extra variable.

[gwilymh]

I think I have found a solution. CuffDiff actually does output a count of each feature (genes.count_tracking, cds.count_tracking, etc). I ran cuffDiff with each sample treated as a separate treatment, c.f. different samples from the same treatment being identified as replicates using the -L/--labels feature. This gave me count data for each feature in each sample.