I am working on a bulk RNA-seq analysis for two groups (disease and control) at three time points: 0h, 24h, 72h. The group samples are paired, with the controls being gene-corrected samples taken from the same individuals. The annotation is as such:
structure(list(group = c("control", "control", "control", "disease",
"disease", "disease", "control", "control", "control", "disease",
"disease", "disease", "control", "control", "control", "disease",
"disease", "disease"), patient = c("P1", "P2", "P3", "P1", "P2",
"P3", "P1", "P2", "P3", "P1", "P2", "P3", "P1", "P2", "P3", "P1",
"P2", "P3"), time = c(0, 0, 0, 0, 0, 0, 24, 24, 24, 24, 24, 24,
72, 72, 72, 72, 72, 72)), row.names = c(NA, -18L), class = "data.frame")
For similar prior analyses I have used the R package TimeSeriesExperiment and gotten good results. However, their method uses an MANOVA approach and is not considering the paired nature of the the two groups. And, in this particular case, there are no replicates per time point, so the patient samples are functioning as per group replicates in the analysis. I have looked around for a method which can account for the pairing, but I'm getting a bit confused over which would be appropriate. Right now, I'm wondering if something like ANCOVA would work, with the pairs as covariate? I am also looking at a limma design, which would be:
(disease_72h - disease_0h) - (control_72h - disease_0h)
... with comparisons for all combinations of time points.
Does anyone have experience with similar analysis/designs and can offer a recommendation, either for an appropriate method or a package which handles such?