I have some single-cell data I want to analyze in DESeq2 with a model of interaction terms of treatment and time variables. See the image for the metadata of samples I have. The data comes from several hashed captures, each with 2 samples. I wanted to model the capture to account for batch effects, but I've tried several permutations and keep getting linear combinations of terms. Is there a way to model both the individual and the capture it came from? Note that the same 2 individuals are always paired in a capture.
My thought for the model was something like this:
~ 0 + cluster_labels + condition:ind.condition.ID + timepoint:cap.timepoint.ID + condition * timepoint
The model works if I don't include the timepoint:cap.timepoint.ID term. Cluster labels isn't shown in the metadata table, but it's the celltypes (CM, TFH, etc.). One capture failed, so there are 6 captures for the 'pre' timepoint and 5 for the 'post'.
