meta.clustering: Clustering of Cell-clusters in the immunoClust-pipeline

In immunoClust: immunoClust - Automated Pipeline for Population Detection in Flow Cytometry

Description

This function provides a direct access to the meta-clustering procedure. The method described and discussed in this manuscript is the EMt-classification (EM-method=20) with the number of events for each cluster as weights. It returns the fitted mixture model parameter in an object of class immunoClust.

Usage

meta.Clustering(P, N, K, W, M, S, label=NULL, I.iter=10, B=500, tol=1e-5,
                bias=0.25, alpha=0.5, EM.method=20,
                norm.method=0, norm.blur=2, norm.minG=10)

Arguments

P	The number of observed parameters for the cell event clusters.
N	The number of cell-clustering experiments.
K	The N-dimensional vector with the numbers of cell event clusters in each experiment. The total number of clusters is totK = sum_{i=1}^K K_i.
W	The totK-dimensional vector with weights of all clusters.
M	The totK x P-dimensional matrix of all cluster means.
S	The totK x P x P-dimensional matrix of all cluster covariance matrices.
label	Optional initial cluster assignment. If label equla NULL all clusters are assigned in one cluster in the initial clustering step.
I.iter	The maximum number of major iteration steps.
B	The totK x P x P-dimensional matrix of all cluster covariance matrices.
tol	The tolerance used to assess the convergence of the EM(t)-algorithms.
bias	The ICL-bias used in the EMt-iteration of the meta-clustering.
alpha	A value between 0 and 1 used to balance the bhattacharrya probabilities calculated with either the full covariance matrices or using only the diagonal elements of it. When working with uncompensated FC data very high correlations between parameters may be observed due to spill over. This leads to a very low bhattacharrya probability for two clusters even if they are located nearby. Using a mixture of the probabilities calculated with the complete covariance matrices and the variance information of each parameter avoids this problem. With a value of alpha=1, only the probabilities with complete covariance matrices are applied. A reasonable value for alpha is 0.5.
EM.method	0 = KL-minimization not weighted 1 = BC-maximization not weighted 10 = BC-maximization weighted 2 = EMt-classification not weighted 20 = EMt-classification weighted
norm.method	Normalization function; see meta.Normalize for details.
norm.blur	For the normalization step the A-posterior probabilites of the cell-clusters belonging to a meta.clusters a used. In order to capture narrow cell-clusters reasonable the co-variance of the cell-clusters is blured for the A-posterior probabilities in the normalization step.
norm.minG	Minimum number of obtained meta-clusters required to process the normalization step in the major iteration loop.

Details

This function is used internally by the meta-clustering procedure meta.process in immunoClust.

Value

The fitted model information in an object of class immunoClust.

Author(s)

Till Sörensen [e-mail]

References

Sörensen, T., Baumgart, S., Durek, P., Grützkau, A. and Häupl, T. immunoClust - an automated analysis pipeline for the identification of immunophenotypic signatures in high-dimensional cytometric datasets. Cytometry A (accepted).

See Also

immunoClust-object, meta.SubClustering, meta.process

Examples

data(dat.exp)
d <- meta.exprs(dat.exp)
res <- meta.Clustering(d$P, d$N, d$K, d$clsEvents, d$M, d$S)

immunoClust documentation built on Nov. 8, 2020, 5:19 p.m.