flowTrans: Optimizing transformations for flow cytometry data.
In flowTrans: Parameter Optimization for Flow Cytometry Data Transformation

Description Usage Arguments Details Value Author(s) References Examples

This is the principal function in the package. The function takes data, a transformation function name, dimension names, and optional preselection parameter as input, and runs the maximum likelihood optimization on the data, performs the transformation, and returns the transformed data together with a list of transformation paramters for each dimension. The optimization is set to maximize the likelihood of the parameters given the data, such that the transformed data is normally distributed. An optional parameter n2f=[T,F] allows an automated preselection of an approximately bivariate normal population via the norm2Filter function. The transformation parameters are then optimized for this preselected region, and finally applied globally to the entire range of data. The optional argument parameters.only=[T,F] specifies whether to return the parameters or the transformed data.

1	flowTrans(dat, fun, dims, n2f,parameters.only)

`dat`	The data to be transformed, should be a `flowFrame` object.
`fun`	A character string naming the transformation function to be applied. Can be one of: `mclMultivBiexp`, `mclMultivBoxCox`,`mclMultivArcSinh`, and `mclMultivLinLog`. The `Multiv` series of functions estimate common transformation parameters for all dimensions.
`dims`	A character vector identifying the dimensions to be transformed.
`n2f`	An optional `TRUE, FALSE` value identifying whether the data should be prefiltered with `norm2Filter` to select a region of the data for which to optimize the transformation. Default is `FALSE`.
`parameters.only`	A logical specifying whether to return only the parameters, and not the transformed data. `mclMultivArcSinh` a, b, c; c=0 `mclMultivBiexp` a, b, c, d, f, w; f=0 `mclMultivBoxCox` theta `mclMultivLinLog` theta

The transformation functions are multivariate, common parameter transformations. The implementation utilizes a look up table to call optimization routines for the different transformations as well as optional flowCore and flowClust transformation implementations to optimize parameters and transform data from a single interface. Parameters are optimized such that the transformed data is multivariate-normal.

if parameters.only = FALSE then:

flowTransResult

A flowTransResult object

If parameters.only=TRUE the returned value will be a vector of common transformation parameters: if mclMultivArcSinh is called, the returned vector will contain a, b, and c=0; if mclMultivBiexp is called, the returned vector will contain a, b, c, d, w; if mclMultivBoxCox is called, the returned vector will contain theta; if mclMultivLinLog is called, the returned vector will contain theta.

Greg Finak <greg.finak@ircm.qc.ca>, Raphael Gottardo <raphael.gottardo@ircm.qc.ca>

Finak G, Perez J M, Weng A, Gottardo R. Optimizing Transformations for Flow Cytometry. (Submitted)

data(GvHD)
result <- flowTrans(GvHD[[1]], "mclMultivArcSinh",
          colnames(GvHD[[1]])[1:2], n2f=FALSE, parameters.only=FALSE)
par(mfrow=c(1,2))
contour(GvHD[[1]][,1:2], main="Untransformed FSC vs SSC")
contour(result$result[,1:2], main="Transformed FSC vs SSC")
summary(result)