generating robust biclusters form the set of biclusters

Description

The package contains a number of functions for computing similarity matrix of the biclusters obtained by a variety of methods, initialization seeds or various parameter settings. It uses biclustering output as generated by biclust or fabia. isa2 package can be used to generate the biclusters as well, however, a prior conversion is needed to a biclust object by using isa2.biclust() function. The matrix is used for the construction of hierarchical tree based on overall similarity, row similarity or column similarity to obtain cut-off points for the similarity metric of choice. Various statistics are output per bicluster set: a number of a given gene(compound) or gene (compound) set has been present in any bicluster of output or per run. After the tree is cut, the robiust or super biclusters are obtained in a form of biclust object, which can further be used for plotting of biclusters. Biclusters are submatrices in the data which satisfy certain conditions of homogeneity. For more details on biclusters and biclustering see Madeira and Oliveira (2004).

Details

Author(s)

Tatsiana Khamiakova <tatsiana.khamiakova@uhasselt.be>

References

Madeira and Oliveira (2004) Biclustering algorithms for biological data analysis: a survey. IEEE/ACM Trans Comput Biol Bioinform. 2004 Jan-Mar;1(1):24-45. Shi et al. (2010) A bi-ordering approach to linking gene expression with clinical annotations in gastric cancer. BMC Bioinformatics. 11. pages 477.

Class BiclustSet

Description

BiclustSet Class contains the biclustering result in a form: bicluster rows and bicluster columns

Objects from the Class

Objects can be created by calls of the form new("BiclustSet", ...). The variety of inputs variety of inputs (isa2, fabia, biclust,...) can be used.

Slots

GenesMembership:

logical, object of class "matrix", with row membership within a bicluster

ColumnMembership:

logical, object of class "matrix", with column membership within a bicluster

Number:

code"numeric", number of biclusters in the set

Author(s)

Tatsiana Khamiakova

Constructor of BiclustSet object

Description

The method extract relevant information from a variety of biclustering input and constructs a BiclustSet object

Methods

signature(x = "ANY")

signature(x = "Biclust")

Converts Biclust objects into BiclustSet object

signature(x = "Factorization")

Converts FABIA Factorization object into BiClustSet

signature(x = "list")

Converts a list with biclustering results into BiClustSet

See Also

BiclustSet

Examples

 test <- matrix(rnorm(5000), 100, 50)
 test[11:20,11:20] <- rnorm(100, 3, 0.1)
 test[17:26,21:30] <- rnorm(100, 3, 0.1)
#Run FABIA
 set.seed(1)
 FabiaRes1 <- fabia(test)
#construct BiclustSet object from FABIA output
 FabiabiclustSet <- BiclustSet(FabiaRes1)
 FabiabiclustSet 

Hierarchical structure of bicluster output

Description

Constructs and plots hierarchical tree of biclusters output based on the similarity matrix

Usage

HCLtree(x)

Arguments

Details

This function operates on a similarity matrix, which is converted to the distance between biclusters according to dist(a,b)= 1-sim(a,b), where the smaller the distance, the higher is overlap in terms of rows and columns. The tree is constructed using complete method and plotted. Further, the structure must be explored and robust or super-biclusters obtained after cutting the tree. identify function can be applied to the hierarchical tree to see the partition and get the plots of biclusters.

Value

tree

Author(s)

Tatsiana Khamiakova

See Also

similarity, plotSuper

Examples

#compute sensitivity for BiMAX biclusters
 test <- matrix(rnorm(5000), 100, 50)
 test[11:20,11:20] <- rnorm(100, 3, 0.1)
 test[17:26,21:30] <- rnorm(100, 3, 0.1)
 testBin <- binarize(test,2)
 res <- biclust(x=testBin, method=BCBimax(), minr=4, minc=4, number=10)
 BiMaxBiclustSet <-  BiclustSet(res)
 SensitivityMatr<- similarity(BiMaxBiclustSet,index="sensitivity")
#construct hierarchical clustering based on the sensitivities
HCLsensitivity <- HCLtree(SensitivityMatr) 
plot(HCLsensitivity, main="structure of bicluster solution")

Combine two Biclust objects into one

Description

Combine two Biclust objects into one

Usage

combine(x,y)

Arguments

Details

If a biclust function returns empty set, joined result contains only results of non-empty object. Info and Parameters slots of a "Biclust" object contain information about both biclustering runs.

Value

object of a class Biclust

Author(s)

Tatsiana Khamiakova

See Also

BiclustSet

Examples

#combine output of two biclust objects
 test <- matrix(rnorm(5000), 100, 50)
 test[11:20,11:20] <- rnorm(100, 3, 0.1)
 test[17:26,21:30] <- rnorm(100, 3, 0.1)
 set.seed(1)
 PlaidRes1 <- biclust(x=test, method=BCPlaid())
 set.seed(2)
 PlaidRes2 <- biclust(x=test, method=BCPlaid())
 combinedRes <- combine(PlaidRes1,PlaidRes2)
 summary(combinedRes)

Get frequency statistic for the columns and rows membership

Description

For a given Bicluster set, for each row and column in data, compute frequency of apperance within a bicluster

Usage

getStats(x)

Arguments

Value

a list of column and row frequencies

Author(s)

Tatsiana Khamiakova

Jaccard similarity Matrix for bicluster output

Description

computes jaccard similarity matrix for biclusters in two bicluster sets

Usage

jaccardMat(x, y, type=c("rows", "cols", "both"))

Arguments

Details

This function operates on BiclustSet objects and computes pairwise similarity based on the common elements between biclusters. The Jaccard similarity score ja for two biclusters A and B is computed as

ja=\frac{|A\cap B|}{|A\cup B|}

Value

matrix of pairwise Jaccard indices

Author(s)

Tatsiana Khamiakova

See Also

similarity,kulczynskiMat, ochiaiMat, sensitivityMat, specificityMat,sorensenMat

Kulczynski similarity Matrix for bicluster output

Description

computes Kulczynski similarity matrix for biclusters in two bicluster sets

Usage

kulczynskiMat(x, y, type=c("rows", "cols", "both"))

Arguments

Details

This function operates on BiclustSet objects and computes pairwise similarity based on the common elements between biclusters. Kulczynski similarity score ku for two biclusters A and B is computed as

ku = 2|A\cap B|\left(\frac{1}{|A|} + \frac{1}{|B|}\right)

Value

matrix of pairwise Kulczynski indices

Author(s)

Tatsiana Khamiakova

See Also

similarity,jaccardMat, ochiaiMat, sensitivityMat, specificityMat,sorensenMat

Ochiai similarity Matrix for bicluster output

Description

Computes Ochiai similarity matrix for biclusters in two bicluster sets

Usage

ochiaiMat(x, y, type=c("rows", "cols", "both"))

Arguments

Details

This function operates on BiclustSet objects and computes pairwise similarity based on the common elements between biclusters. The Ochiai similarity score ja for two biclusters A and B is computed as

oc=\frac{|A\cap B|}{\sqrt{|A| |B|}}

Value

matrix of pairwise Ochiai indices

Author(s)

Tatsiana Khamiakova

See Also

similarity,jaccardMat, ochiaiMat, sensitivityMat, specificityMat,sorensenMat

Plot Gene Expression Profiles Across All Samples of the Original Data

Description

Plot Gene Expression Profiles Across All Samples of the Original Data

Usage

plotProfilesAcrossAllSamples(x, coreBiclusterGenes, coreBiclusterSamples)

Arguments

Details

The plot re-sorts the samples by bicluster membership and highlights them in red. Only the genes of a bicluster are plotted.

Value

no return value; a plot is drawn to the current device

Author(s)

Tatsiana Khamiakova

See Also

BiclustSet ,plotSuperAll

Plot Gene Expression Profiles within a (Core) Bicluster

Description

Plot Gene Expression Profiles within a (Core) Bicluster

Usage

plotProfilesWithinBicluster(x, main = "", sampleNames, geneNames = NULL)

Arguments

Value

no return value; a plot is drawn to the current device

Author(s)

Tatsiana Khamiakova

See Also

plotSuper

Plot gene profiles within biclusters

Description

Function for plotting gene profiles for compounds within constructed super-bicluster

Usage

plotSuper(x, data, BiclustSet)

Arguments

Details

This function constructs a robust bicluster from a set of biclusters identified in a hierarchical tree and plots gene profiles for columns in a robust bicluster. Each line represents a gene from a bicluster. The bicluster is saved as Biclust object which can be further plotetd by available functions from biclust package. The information about the number of biclusters used to generate the resulting robust bicluster is saved in Call slot of the object. This information is important to see how often the bicluster has been discovered under different parameter settings (e.g. initialization seeds) Indices used as an input can be obtained by identify function or by cutting the tree.

Value

biclust object containing bicluster and the information about bicluster subset used to generate it

Author(s)

Tatsiana Khamiakova

See Also

HCLtree, plotSuperAll, plotProfilesWithinBicluster

Plot gene profiles for all samples in the data

Description

Function for plotting bicluster gene profiles for all samples in the data

Usage

plotSuperAll(x, data, BiclustSet)

Arguments

Details

This function constructs a robust bicluster from a subset of biclusters specified in x argument and plots the expression profiles

Value

biclust object

Author(s)

Tatsiana Khamiakova

See Also

HCLtree, plotProfilesAcrossAllSamples

Sensitivity Matrix for bicluster output

Description

Computes sensitivity matrix for biclusters in two bicluster sets

Usage

sensitivityMat(x, y, type=c("rows", "cols", "both"))

Arguments

Details

This function operates on BiclustSet objects and computes pairwise similarity based on the common elements between biclusters. Sensitivity inclusion score sen of biclusters A and B is computed as

sen=\frac{|A\cap B|}{|A|}

Value

matrix of pairwise sensitivities

Author(s)

Tatsiana Khamiakova

See Also

similarity,jaccardMat, ochiaiMat, kulczynskiMat, specificityMat,sorensenMat

Similarity Matrix for bicluster output

Description

computes similarity matrix for the biclustering output based on one of the pairwise similarity indices of biclusters in a given bicluster set

Usage

similarity(x, index = "jaccard", type="rows")

Arguments

Details

This function operates on BiclustSet object and computes pairwise similarity based on the common elements between biclusters. type variable controls whether similarity index is constructed for all elements, or in one dimension (rows or columns) only. In general, similarity indices for one dimension (row or column) are higher than for two-dimensions. Several options for similarity indices are offered: jaccard (default), kulczynski, sensitivity, specificity, sorensen and ochiai indices.

Value

a "similarity" object containing similarity matrix

Author(s)

Tatsiana Khamiakova

See Also

HCLtree, plotSuper , jaccardMat,kulczynskiMat, ochiaiMat, sensitivityMat, specificityMat,sorensenMat

Examples

#compute sensitivity for BiMAX biclusters
 test <- matrix(rnorm(5000), 100, 50)
 test[11:20,11:20] <- rnorm(100, 3, 0.1)
 test[17:26,21:30] <- rnorm(100, 3, 0.1)
 testBin <- binarize(test,2)
 res <- biclust(x=testBin, method=BCBimax(), minr=4, minc=4, number=10)
 BiMaxBiclustSet <-  BiclustSet(res)
 SensitivityMatr<- similarity(BiMaxBiclustSet,index="sensitivity", type="rows")
 SensitivityMatr

Sorensen similarity Matrix for bicluster output

Description

Computes Sorensen similarity matrix for biclusters in two bicluster sets

Usage

sorensenMat(x, y, type=c("rows", "cols", "both"))

Arguments

Details

This function operates on BiclustSet objects and computes pairwise similarity based on the common elements between biclusters. Sorensen similarity score so for two biclusters A and B is computed as

so = \frac{2|A\cap B|}{|A| + |B|}

Value

matrix of pairwise Sorensen indices

Author(s)

Tatsiana Khamiakova

See Also

similarity,jaccardMat, ochiaiMat, sensitivityMat, specificityMat,kulczynskiMat

Specificity Matrix for bicluster output

Description

Computes specificity matrix for biclusters in two bicluster sets

Usage

specificityMat(x, y, type=c("rows", "cols", "both"))

Arguments

Details

This function operates on BiclustSet objects and computes pairwise similarity based on the common elements between biclusters. Sensitivity inclusion score spe of biclusters A and B is computed as

spe=\frac{|A\cap B|}{|B|}

Value

matrix of pairwise specificities

Author(s)

Tatsiana Khamiakova

See Also

similarity,jaccardMat, ochiaiMat, kulczynskiMat, sensitivityMat,sorensenMat