cluster: Wrapper to clustering algorithms

In UCLA-BD2K/CV.Signature.TCP: Cardiovascular Signature Temporal Clustering Platform

Description

This is a convenient wrapper to Kmeans and Hierarchical clustering when using the 'tms' package.

Usage

cluster(
  dat,
  timepoints = NULL,
  K,
  dist.method = c("euclidean", "cor.diss", "dtw"),
  cluster.method = c("kmeans", "hclust"),
  hclust.algorithm = "complete",
  kmeans.algorithm = "Hartigan-Wong",
  kmeans.centers,
  center.dat = TRUE,
  scale.dat = FALSE,
  verbose = FALSE,
  seed = NULL,
  ...
)

Arguments

dat	a (denoised) data matrix with m biomarkers as rows, over n time points (columns).
timepoints	a vector of time points for columns of dat. While this is not necessary, it's here for now to remind the user about the analysis pipeline.
K	a number of clusters.
dist.method	a distance method for time course data, resulting in a m * m distance matrix for rows. 'dtw' for dynamic time wrapping or 'cor.diss' for correlation-based dissimilarities.
cluster.method	a clustering method.
hclust.algorithm	an algorithm used in hierarchical clustering. See hclust for a full list.
kmeans.algorithm	an algorithm used in kmeans clustering. See kmeans for a full list.
center.dat	a logical specifying to center the input and denoised data. By default, TRUE.
scale.dat	a logical specifying to scale the input and denoised data. By default, FALSE.
verbose	a logical specifying to print the computational progress. By default, FALSE.
seed	a seed for the random number generator.
...	optional arguments.

Value

cluster returns a list consisting of

dat.dtw	m * m distance matrix based on dynamic time wrapping.
cluster.obj	an object returned from clustering the predicted data.
membership	a vector of length m, identities of clusters.

Author(s)

Neo Christopher Chung [e-mail]

Examples

## Not run: 
## load the example "coptm" data
data(cys_optm)
meta <- cys_optm[,1:4]
optm <- log(cys_optm[meta$Select,5:10])
optm <- t(scale(t(optm), scale=TRUE, center=TRUE))
days <- as.numeric(colnames(optm))

## denoise using the cubic splines
denoised_optm <- denoise_spline(optm, timepoints = days, dof="cv", verbose=FALSE)

## cluster the denoised data using K-means clustering
clustered_optm <- cluster(denoised_optm,
                timepoints = days,
                cluster.method = "kmeans",
                K=6,
                center.dat = TRUE,
                scale.dat = FALSE,
                verbose = TRUE)

## End(Not run)

UCLA-BD2K/CV.Signature.TCP documentation built on May 15, 2020, 11:27 p.m.