Cyclic Loess normalization for Hi-C data

cyclic_loess(
  hicexp,
  iterations = 3,
  span = NA,
  parallel = FALSE,
  verbose = FALSE
)

Arguments

hicexp

A hicexp object

iterations

The number of iterations (cycles) of loess normalization to perform. Defaults to 3.

span

The span for loess normalization. Defaults to NA indicating that span will be automatically calculated using generalized cross validation.

parallel

Logical. Should parallel processing be used?

verbose

Logical. Should messages about loess normalization be printed to the screen.

Value

A hicexp object that has been normalized.

Details

This function performs cyclic loess normalization on a Hi-C experiment. multiHiCcompare's cyclic loess procedure is a modified version of Ballman's (2004) cyclic loess and the joint loess normalization used in the original HiCcompare. For each unique pair of samples in the hicexp object an MD plot is generated. A loess curve is fit to the MD plot and then the fitted values are used to adjust the data. This is performed on all unique pairs and then repeated until convergence.

Examples

#' data("hicexp2") hicexp2 <- cyclic_loess(hicexp2, span = 0.7)