cellranger: min gene count and UMI/cell filter?

Question

[mcsimenc]

I'm trying to re-analyze a 10x scRNA-seq dataset and follow the methods detailed in the associated publication (main text & supplement) but I have been unable to find the cellranger functionality they say they used: filtering cellranger aggr output by minimum number of genes per cell and minimum UMIs per cell. Here is the portion from the supplement where they explain what they did:

"Initial filtering in CellRanger (gene count > 200), recovered 15,918 cells. To ensure we only use high-quality cells for further analysis, we used the filtered data provided by cellranger. This corresponds to a minimum UMI count of 17,290 in the aggregated data, recovering a final 5,145 cells."

From my understanding the filtered data from cellranger count and aggr includes only barcodes which were assigned to cells but does not filter based on gene or UMI count. It's totally possible I have simply overlooked an explanation of options to do this in the cellranger documentation, but I have been unable to find any such options so far.

Does anyone know how I can use cellranger to filter cells based on minimum gene and UMI count?

Thanks!

Matt

Answer 1

[swbarnes2]

I don't see a way to do that filtering with cellranger, but it's very easy to do in Seurat. Maybe it's a mistake in their description.

The decision as to whether or not a cell barcode represents a real cell or just cruft should be based on UMI count, but cellranger handles that. They might just be reporting empirically what the lowest UMI count is in their filtered cells, not saying they personally imposed that as a cutoff.