CAMMGCluster: MG cluster detection for CAM

Description Usage Arguments Details Value Examples

Description

This function finds corner clusters as MG clusters (clusters containing marker genes).

Usage

1
CAMMGCluster(K, PrepResult, generalNMF = FALSE, nComb = 200)

Arguments

K

The candidate subpopulation number.

PrepResult

An object of class "CAMPrepObj" obtained from CAMPrep function.

generalNMF

If TRUE, the decomposed proportion matrix has no sum-to-one constraint for each row. Without this constraint, the scale ambiguity of corner cluster centers will not be removed when computing reconstruction errors. The default is FALSE.

nComb

The number of possible combinations of clusters as corner clusters. Within these possible combinations ranked by margin errors, we can further select the best one based on reconstruction errors. The default is 200.

Details

This function is used internally by CAM function to detect clusters containing marker genes, or used when you want to perform CAM step by step.

This function provides two solutions. The first is the combination of clusters yielding the minimum sum of margin-of-errors for cluster centers. In the second, nComb possible combinations are selected by ranking sum of margin-of-errors for cluster centers. Then the best one is selected based on reconstruction errors of all data points in original space.

Value

An object of class "CAMMGObj" containing the following components:

idx

Two numbers which are two solutions' ranks by sum of margin-of-error.

corner

The indexes of clusters as detected corners. Each row is a solution.

error

Two rows. The first row is sum of margin-of-errors for nComb possible combinations. The second row is reconstruction errors for nComb possible combinations.

Examples

1
2
3
4
5
6
7
8
9
#obtain data
data(ratMix3)
data <- ratMix3$X

#preprocess data
rPrep <- CAMPrep(data, dim.rdc = 3, thres.low = 0.30, thres.high = 0.95)

#Marker gene cluster detection with a fixed K = 3
rMGC <- CAMMGCluster(3, rPrep)

Lululuella/CAMTHC documentation built on May 5, 2019, 2:39 a.m.