Finds barcode count cutoff point

Source: R/AllGenerics.R, R/method-cure_barcode.R

Finds the cutoff point for the barcode count filtering based on the barcode count distribution.

bc_auto_cutoff(barcodeObj, useCleanBc = TRUE)

# S4 method for class 'BarcodeObj'
bc_auto_cutoff(barcodeObj, useCleanBc = TRUE)

Arguments

barcodeObj

A BarcodeObj object.

useCleanBc

A logical value, if TRUE, the cleanBc slot in the BarcodeObj object will be used, otherwise the messyBc slot will be used.

Value

a numeric vector of the cutoff point.

Details

The one dimension kmeans clustering is applied to identify the "true barcode" based on the read count. The algorithm detail is: 1. Remove the barcodes with counts below the median of counts. 2. Transform the count by log2(x+1). 3. Apply the 1-dimension clustering to the log count, with the cluster number of 2 and weights of the log count. 4. Choose the minimum count value in the cluster with more counts as cutoff point.

For more info about 1 dimension kmeans used here please refer to Ckmeans.1d.dp.

Examples


data(bc_obj)

bc_auto_cutoff(bc_obj)
#> ------------
#> bc_auto_cutoff: cleanBc is used.
#> ------------
#> test1 test2 
#>   104    53