Hi

I have 9 time points (so CMO301 means tag of time point one and so on) from the same participant processed in two different ways and I get very different cell counts per time point.

When I analyse each time point separately using filtered_feature_bc_matrix from per-time-point Cell Ranger runs, I get ~800–1100 cells per time point (TP1–TP9).

table(data$TimePoint)

 TP1  TP2  TP3  TP4  TP5  TP6  TP7  TP8  TP9 
1094  977  964  804  930 1015 1004 1025 1111

When I analyse the pooled data using filtered_feature_bc_matrix and Seurat hashing (HTODemux), most cells are classified as Negative or Doublet, and only a few dozen cells are assigned to each CMO/time point.

table(data$hash.ID)

 Doublet Negative   CMO301   CMO305   CMO303   CMO302   CMO304   CMO308   CMO307 
     706      235       37       20       24       14       13       13       11 
  CMO309   CMO306 
      12        9

Why do these two approaches give such different numbers of cells per time point? Is this discrepancy expected with CellPlex/HTO data, and which counts should be considered reliable for downstream time-resolved analysis in Seurat?

Thanks for any idea