Cell Filtering

DRAGEN uses a UMI count threshold to separate cell-barcodes corresponding to single-cells in the original sample from background noise. Cell-barcodes with UMI counts beneath the threshold are considered background noise. To set the threshold, DRAGEN sorts all cell-barcodes by their UMI count in descending order and then determines the counts of the following two reference cell-barcodes:

•

topCount—The maximum corresponding to the count of the cell-barcode at the first percentile of the UMI count distribution.

•

minCount—An expected minimum corresponding to the count of the last expected cell. The minCount is represented as the _N_th cell, where N is the number of expected cells.

The final threshold is then set as max(topCount * 0.1, minCount * 0.5.