Calculate the Dispersion Separability Criterion (DSC)

This function computes the Dispersion Separability Criterion (DSC) metric which was defined by the MD Anderson Bioinformatics team for computing the strength of batch effects in high-throughput sequencing data from TCGA.

Usage

dsc(x, g)

Arguments

x

A numeric matrix (features in rows, samples in columns)

g

A vector containing the batch assignment for each sample

Value

DSC metric

Examples

set.seed(123)

# Small DSC when samples drawn from same distribution
x1 <- matrix(rnorm(500 * 100), nrow = 500, ncol = 100)
batches <- rep(c("Batch_1", "Batch_2"), each = 50)
dsc(x = x1, g = batches)
#> [1] 0.09999285

# Large DSC when samples have batch effect present
x2 <- x1
x2[, 51:100] <- x2[, 51:100] + 50
dsc(x2, batches)
#> [1] 25.24697