flowStats (version 3.30.0)

idFeaturesByBackgating: (Internal use only) Identify features of flow cytometry data using backgating

Description

Identify and labeling significant features using divisive clustering method such as diana.

Usage

idFeaturesByBackgating(bg, nDim, thres.sigma=2.5, lambda=0.1,
                       reference.method="median",
                       plot.workflow=FALSE, ask=names(dev.cur())!="pdf")

Arguments

bg

A data frame containing subpopulations on channels of interests. Must be a returning result from flowStats:::backGating

nDim

An integer indicating the length of channels of interest.

thres.sigma

An numerical value indicating the threshold at which to cut tree, e.g., as resulting from 'diana', into several clusters.

lambda

A numerical value indicating the percentage of the potential features that is used as a threshold for deciding outlier clusters. The default value is 0.1.

reference.method

A character vector indicating the method for computing the reference features. If median, the reference feature is defined by the medain of eac cluster of features. Valid methods include median and mean only.

plot.workflow

Logical. If TURE, display the workflow of feature identification.

ask

Logical. If TRUE, the display operates in interactive mode.

Value

• register
  A list containing registered features for each sample.

Details

Using the resulting data frame from backGating as potential features, the algorithm follows four major steps: (i) centering the potential features, which yields the returning value TransMatrix, (ii) using diana to compute a clustering of the potential features, (iii) cutting the tree into several clusters, and (iv) accessing outliers and rendering the final registered features with labels.

In step three, the threshold for cutting the tree is computed by $$sd * thres.sigma,$$ where $sd$ is the standard deviation of the distribution of the height between entities computed by diana.

A cluster is determined as an outlier if the number of its members is less than the median of the numbers of all clusters' members times 'lambda'.

See Also

diana, BackGating, gpaSet.

Examples

data(ITN)
wf <- workFlow(ITN)
tl <- transformList(colnames(ITN)[3:7], asinh, transformationId="asinh")
dat <- trnasformList(ITN, tl)
bg <- backGating(dat, xy=c("FSC", "SSC"), channels="CD3")

data(BackGating)
results <- flowStats:::idFeaturesByBackgating(bg=BackGating, nDim=2,
                       plot.workflow=TRUE, ask=TRUE)