Dear dear all,

Here is my study design.

I have three groups of experiments: (1) tumor_adj (n=3); (2) non_tumor (n=3); and (3) control groups (n=2).

There are two repeated measures (timepoints) per group: (i) pre, and (ii) post.

Plus, I have an additional cancer (n=3) group ONLY at the post timepoint.

I have tried

to use m1 <- model.matrix(~ condition, source) to remove the "cancer:pre" from the matrix cause it does not exist. Then I supply the matrix to dds with design(dds) <- m1.

Question:

1. Do I need to do anything with my n numbers per group? They are my biological replicates.

2. How would you model my experiments?

3. Should I remove the additional cancer group so that all others fit the general model? But I still want to normalize my raw count matrix at once...

4. How do I achieve the following:

   • 4.1. Test the differences between matched groups, including (i) the main effect of "Group" and (ii) perform pairwise comparisons between groups at each time point;
   • 4.2. Test the temporal changes within groups, including (i) the main effect of "Time" and (ii) perform pairwise comparisons between timepoints within group;
   • 4.3. Test the interaction term "Group:Time" to assess whether temporal changes differ between the three groups.

5. Am I doing correctly in the step of supplying my own matrix? If I want to test the complex interactions, how should I set the factors and matrix??

6. I have difficulty in generating the results. In particular, why is there only one p-value? I have multiple groups comparison. How should I do it properly if I want to have the normalized counts, and all 13 comparison p-values (padj) listed together in the result table.

Side question:

Should I be using VST-lmerSeq pipeline to capture the temporal changes? And how should I do it? If anyone have experience, please help....

Many thanks and Merry Christmas!!!

Jimmy