Can I use combat on circadian RNA-seq timeseries

[alex297]

Hello,
We have mouse RNA-seq with 6 time points over 24 h period and 3 replicates for each time points (18 samples total), each sample is from different mouse, all measurements were performed in one experiment. PCA plot on all genes (below) shows that there are substantial variations between samples (often larger than between time points). I think that these are biological variations, and circadian genes show robust rhythm, not much noise. I used BooteJTK to detect rhythmic genes and generally satisfied with results, even though we found only 117 significantly rhythmic genes. But the low number of rhythmic genes is probably related with sparse data (only 1 day measurements with 4h intervals).
However, my boss asked me to perform batch correction to group samples belonging to the same time point together. Would this be appropriate to use Combat? I saw some warnings that it should be used only when there are clear batches, but in our case each sample comes from different mouse, so I am not sure whether we have any batches.
I would be grateful for any advice/comment.
Best regards
Alex