DESeq2 replicates in time course data

Question

[fernando.perez8]

Hello,

I am analyzing data from an RNAseq experiment in which there are 4 control and 4 treatment individuals. From each individual a sample was taken at time 0, 12 and 24 hours. That is, 8 individuals, 2 conditions, 3 samples per individual: total 4x2x3 = 24 samples. I want to do a simple comparison of control vs. treatment. I wonder if I should treat the different samples from the same individual as technical replicates and therefore should be collapsed or consider them as biological replicates despite being from the same individual, since they have been taken at different times. In this particular case I am not interested in considering time as a factor.

thank you very much for your help

Here is my colData

   sampleid condition time
1      ind1      ctrl    0
2      ind1      ctrl   12
3      ind1      ctrl   24
4      ind2      ctrl    0
5      ind2      ctrl   12
6      ind2      ctrl   24
7      ind3      ctrl    0
8      ind3      ctrl   12
9      ind3      ctrl   24
10     ind4      ctrl    0
11     ind4      ctrl   12
12     ind4      ctrl   24
13     ind5     treat    0
14     ind5     treat   12
15     ind5     treat   24
16     ind6     treat    0
17     ind6     treat   12
18     ind6     treat   24
19     ind7     treat    0
20     ind7     treat   12
21     ind7     treat   24
22     ind8     treat    0
23     ind8     treat   12
24     ind8     treat   24

Answer 1

[Carlo Yague]

Even if you do not want to explicitely test the effect of time on gene expression, it is recommanded to include in the DESeq2 model all the terms that you think might affect gene expression. In such a case, 'time' will be seen as a covariate, similar to the "batch" factor that we see in many examples of DESeq2 models.

So, no, unless you have good reason to think that "time" does not have any effect on gene expression, don't use the different times as replicates (they are not).

[fernando.perez8]

Thanks you for your reply. I agree that different times are not replicas, but I have seen some publications (e.g. this ) where they treat them as technical replicas and I was already a bit confused.