Pseudotime analysis on multiple samples

Question

I am using Monocle3 to run pseudo time analysis on samples at two different conditions (samples are paired). My question is, is it valid to combine all the cells from samples at specific condition when clustering, learning graph and ordering cells? or should I pick up only one sample per condition and run the analysis on?

My understanding is that pseudotime analysis uses the expression data of cells to try to infer the trajectory of cells, so I believe it might not be correct when I pool all the cells that might not have a true trajectory?

Thanks

Answer 1

I would try it both ways (i.e. run pseudotime analysis on individual and combined samples). If you have the same transcriptional trajectories in multiple samples then you might be able to capture this via combining samples, and arguably the most interesting signatures are generalizable across samples. On the other hand, there might be batch effects that disrupt the signal of a combined analysis.

I think I understand your concern is that different samples have independent cells and so they should not form part of one trajectory. However, even within a sample, we are not capturing the true trajectory as this would require sampling the same individual cell multiple times at different time points, which is not currently possible because scRNA-seq requires lysing the cells (see also live-seq).