Determines an optimal weight for weighted clustering by similarity weighted clustering.

Share:

Description

The function DetermineWeight_SimClust determines an optimal weight for performing weighted similarity clustering on by applying similarity clustering. For each given weight, is each separate clustering compared to the clustering on a weighted dissimilarity matrix and a Jaccard coefficient is calculated. The ratio of the Jaccard coefficients closets to one indicates an optimal weight.

Usage

1
2
3
4
5
DetermineWeight_SimClust(List, type = c("data","dist","clusters"),
weight=seq(0, 1, by = 0.01),nrclusters = NULL, distmeasure = c("tanimoto",
"tanimoto"),normalize=FALSE,method=NULL,clust = "agnes",linkage=c("ward","ward")
,alpha=0.625,gap = FALSE, maxK = 50, names = c("B", "FP"),StopRange=FALSE,
plottype="new",location=NULL)

Arguments

List

A list of matrices of the same type. It is assumed the rows are corresponding with the objects.

type

Type indicates whether the provided matrices in "List" are either data matrices, distance matrices or clustering results obtained from the data. If type="dist" the calculation of the distance matrices is skipped and if type="clusters" the single source clustering is skipped. Type should be one of "data", "dist" or"clusters".

weight

Optional. A list of different weight combinations for the data sets in List. If NULL, the weights are determined to be rqual for each data set. It is further possible to fix weights for some data matrices and to let it vary randomly for the remaining data sets. An example is provided in the details.

nrclusters

The number of clusters to cut the dendrogram in. This is necessary for the computation of the Jaccard coefficient.

distmeasure

A character vector with the distance measure for each data matrix. Should be one of "tanimoto", "euclidean", "jaccard","hamming".

normalize

Logical. Indicates whether to normalize the distance matrices or not. This is recommended if different distance types are used. More details on standardization in Normalization.

method

A method of normalization. Should be one of "Quantile","Fisher-Yates", "standardize","Range" or any of the first letters of these names.

clust

Choice of clustering function (character). Defaults to "agnes".

linkage

A vector with the choice of inter group dissimilarity (character) for each data set.

alpha

The parameter alpha to be used in the "flexible" linkage of the agnes function. Defaults to 0.625 and is only used if the linkage is set to "flexible"

gap

Logical. Whether or not to calculate the gap statistic in the clustering on each data matrix separately. Only if type="data".

maxK

The maximal number of clusters to consider in calculating the gap statistic. Only if type="data".

names

The labels to give to the elements in List.

StopRange

Logical. Indicates whether the distance matrices with values not between zero and one should be standardized to have so. If FALSE the range normalization is performed. See Normalization. If TRUE, the distance matrices are not changed. This is recommended if different types of data are used such that these are comparable.

plottype

Should be one of "pdf","new" or "sweave". If "pdf", a location should be provided in "location" and the figure is saved there. If "new" a new graphic device is opened and if "sweave", the figure is made compatible to appear in a sweave or knitr document, i.e. no new device is opened and the plot appears in the current device or document.

location

If plottype is "pdf", a location should be provided in "location" and the figure is saved there.

Details

If the type of List is data, an hierarchical clustering is performed on each data matrix separately. After obtaining clustering results for the two data matrices, the distance matrices are extracted. If these are not calculated with the same distance measure, they are normalized to be in the same range. For each weight, a weighted linear combination of the distance matrices is taken and hierarchical clustering is performed once again. The resulting clustering is compared to each of the separate clustering results and a Jaccard coefficient is computed. The ratio of the Jaccard coefficients closets to one, indicates an optimal weight. A plot of all the ratios is produced with an extra indication for the optimal weight.

The weight combinations should be provided as elements in a list. For three data matrices an example could be: weights=list(c(0.5,0.2,0.3),c(0.1,0.5,0.4)). To provide a fixed weight for some data sets and let it vary randomly for others, the element "x" indicates a free parameter. An example is weights=list(c(0.7,"x","x")). The weight 0.7 is now fixed for the first data matrix while the remaining 0.3 weight will be divided over the other two data sets. This implies that every combination of the sequence from 0 to 0.3 with steps of 0.1 will be reported and clustering will be performed for each.

Value

The returned value is a list with three elements:

ClustSep

The result of Cluster for each single element of List

Result

A data frame with the Jaccard coefficients and their ratios for each weight

Weight

The optimal weight

Author(s)

Marijke Van Moerbeke

References

PERUALILA-TAN, N., SHKEDY, Z., TALLOEN, W., GOEHLMANN, H. W. H., QSTAR Consortium, VAN MOERBEKE, M., KASIM, A., (in press). Weighted-Similarity Based Clustering of Chemical Structure and Bioactivity Data in Early Drug Discovery. Journal of Bioinformatics and Computational Biology.

See Also

WeightedSimClust

Examples

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
data(fingerprintMat)
data(targetMat)

MCF7_F = Cluster(fingerprintMat,type="data",distmeasure="tanimoto",normalize=FALSE,
method=NULL,clust="agnes",linkage="ward",alpha=0.625,gap=FALSE,maxK=55,StopRange=FALSE)
MCF7_T = Cluster(targetMat,type="data",distmeasure="tanimoto",normalize=FALSE,
method=NULL,clust="agnes",linkage="ward",alpha=0.625,gap=FALSE,maxK=55,StopRange=FALSE)

L=list(MCF7_F,MCF7_T)

MCF7_Weight=DetermineWeight_SimClust(List=L,type="clusters",weight=seq(0,1,by=0.01),nrclusters=7,
distmeasure=c("tanimoto","tanimoto"),normalize=FALSE,method=NULL,clust="agnes",
linkage=c("flexible","flexible"),alpha=0.625,gap=FALSE,maxK=50,names=c("FP","TP"),StopRange=FALSE,
plottype="new",location=NULL)

Want to suggest features or report bugs for rdrr.io? Use the GitHub issue tracker.