yaConsensus (version 1.1)

yaConsensus: yaConsensus computes a number of hierarchical clusterings by sampling either samples or features.

Description

This function mainly generates a list of "hclust" objects for downstream analysis.

Usage

yaConsensus(ddata, runs = 1000, epsilon = 0.65, is_by_sample = TRUE, 
            distMethod = "euclidean", hcMethod = "ward.D2", prefix = NULL)

Value

A named list with the following slots:

distMethod

matches the input

hcMethod

matches the input

lables

a string list with the names of the samples

bySample

matches 'is_by_sample' input parameter

epsilon

matches the input

subsetDimension

actual dimension of the subsets

runs

matches the input

hclust

a list of 'hclust' objects

elapsed_time

time (in seconds) required

ncores

the number of cores used

Arguments

ddata

either a data matrix (samples in rows, and features in columns), or a "dist" object.

runs

an integer value for the number of samplings.

epsilon

a real value indicating the sampling rate.

is_by_sample

a logical value indicating if the sampling is by samples (TRUE) or features (FALSE).

distMethod

a character indicating the kind of distance for the inner clustering. It can be any of the methods from the dist function.

hcMethod

a character indicating the linkage mathod of the inner clustering. It can be any of the methods from the hclust function.

prefix

string specifying a prefix to store the results in a .RData file.

Author

Stefano M. Pagnotta

Details

This function can run sequentially or in parallel. In this case, it is necessary to register a cluster of CPUs according to doParallel protocol.

To get the consensus clustering, the output of the function has to be processed with the plot() function. The consensus dissimilarity follows from the algorithm of Monti et al. (2003). The consensus clustering is from a hierarchical procedure (hclust) with "complete" linkage (outer hc method).

References

Monti et al. (2003) - Consensus Clustering: A Resampling-Based Method for Class Discovery and Visualization of Gene Expression Microarray Data - Machine Learning 52(1-2):91-118 <DOI: 10.1023/A:1023949509487>

See Also

dist, hclust, doParallel::doParallel()

Examples

## Generate data and annotation
n <- 50; m <- 3000
ddata <- matrix(rnorm(n * m), ncol = m)  
ddata[1:20, ] <- ddata[1:20, ] + 0.2
row.names(ddata) <- c(paste0("A", 1:20), paste0("B", 1:30))
ddist <- dist(ddata)

annotation <- data.frame(row.names = rownames(ddata), clust = substr(rownames(ddata), 1, 1))
annotation.colorCode <- c("red", "blue")
names(annotation.colorCode) <- c("A", "B")

####### run in sequential mode
####### sampling the samples ....
(aConsensus <- yaConsensus(ddist))
plot(aConsensus, G = 2)

ans <- plot(aConsensus, G = 2, 
            annotation = annotation, 
            annotation.colorCode = annotation.colorCode)
summary(ans)
summary(ans, given = "clust")

####### sampling the features ....
(aConsensus <- yaConsensus(ddata, runs= 20, epsilon = 0.2, is_by_sample = FALSE))
ans <- plot(aConsensus, G = 2, 
            annotation = annotation, 
            annotation.colorCode = annotation.colorCode,
            matching_clustering = "clust")

summary(ans, given = "clust")


####### run in parallel mode
## uncomment to run

# require(doParallel)
# cpu_cluster <- makeCluster(3)
# registerDoParallel(cpu_cluster)

# (aConsensus <- yaConsensus(ddist))
# plot(aConsensus, G = 2)

#stopCluster(cpu_cluster)