kebabs (version 1.2.3)

heatmap,PredictionProfile,missing-method: Heatmap Methods

Description

Create a heat map of prediction profiles

Usage

"heatmap"(x, Rowv = TRUE, add.expr, margins = c(5, 5), RowSideColors = NULL, cexRow = max(min(35/nrow(x@profiles), 1), 0.1), cexCol = max(min(35/ncol(x@profiles), 1), 0.1), main = NULL, dendScale = 1, barScale = 1, startPos = 1, endPos = ncol(x@profiles), labels = NULL, windowSize = 1, ...)

Arguments

x

prediction profile of class PredictionProfile.

Rowv

determines the row order of the plot. When set to TRUE the profile rows are clustered via hierarchical clustering and a row dendrogram is plotted. When set to FALSE, NA or NULL the order is corresponds to the order of the sequences in the profile. If this parameter has a value of random rows are ordered randomly, for decision the ordering is according to decreasing decision values. A user-defined order can be specified through a numeric vector of indices. Default=TRUE

add.expr

largely analogous to the standard heatmap function.

margins

largely analogous to the standard heatmap function. Default=c(5,5)

RowSideColors

a vector of color values specifying the colors for the side bar. Default=NULL

cexRow

largely analogous to the standard heatmap function. When set to 0 the row labels are suppressed. Default=defined dependent on number of profile rows

cexCol

largely analogous to the standard heatmap function. When set to 0 the column labels are suppressed. Default=defined dependent on number of profile columns

main

largely analogous to the standard heatmap function.

dendScale

factor scaling the width of the row dendrogram; values have to be larger than 0 and not larger than 2. Default=1

barScale

factor scaling the width of the label color bar. Values have to be larger than 0 and not larger than 4. Default=1

startPos

start sequence position. Together with the parameter endPos a subset of sequence positions can be selected for the heatmap. Default=1

endPos

end sequence position (see also startPos). Default=maximum sequence length in the profile.

labels

a numeric vector, character vector or factor specifying the labels for the sequences in the profile. If this parameter is different from NULL the labels are plotted as side bar using the colors specified in the parameter RowSideColors. Default=NULL

windowSize

numerical value specifying the window size of an optional sliding window averaging of the prediction profiles. The value must be larger than 0. Even values are changed internally to odd values by adding 1. Default=1

...

additional parameters which are passed to the image method transparently.

Value

Invisibly, a cluster dendrogram.

Details

The heatmap function provides plotting of heatmaps from prediction profiles with various possibilities for sample (=row) ordering (see parameter Rowv). The heatmap is shown together with an optional color sidebar showing the labels and an optional row cluster dendrogram when hierarchical clustering defines the row order. For long sequences the heatmap can be restricted to a subset of positions. Additionally smoothing can be applied to the prediction profiles through sliding window averaging. Through smoothing important regions can become better visible.

References

http://www.bioinf.jku.at/software/kebabs (Bodenhofer, 2009) -- U. Bodenhofer, K. Schwarzbauer, M. Ionescu and S. Hochreiter. Modelling position specificity in sequence kernels by fuzzy equivalence relations. (Mahrenholz, 2011) -- C.C. Mahrenholz, I.G. Abfalter, U. Bodenhofer, R. Volkmer and S. Hochreiter. Complex networks govern coiled-coil oligomerizations - predicting and profiling by means of a machine learning approach. J. Palme, S. Hochreiter, and U. Bodenhofer (2015) KeBABS: an R package for kernel-based analysis of biological sequences. Bioinformatics (accepted). DOI: 10.1093/bioinformatics/btv176.

See Also

getPredictionProfile

Examples

## load coiled coil data
data(CCoil)

## define annotation specific gappy pair kernel
gappya <- gappyPairKernel(k=1,m=11, annSpec=TRUE)

## train model
model <- kbsvm(x=ccseq, y=as.numeric(yCC), kernel=gappya,
               pkg="e1071", svm="C-svc", cost=15)

## generate prediction profiles
predProf <- getPredictionProfile(ccseq, gappya,
                       featureWeights(model), modelOffset(model))

## show prediction profiles
predProf

## Not run: 
# ## plot heatmap for the prediction profiles - random ordering of samples
# heatmap(predProf, Rowv="random", main="Prediction Profiles", labels=yCC,
# RowSideColors=c("blue", "red"), cexRow=0.15, cexCol=0.3)
# 
# ## plot heatmap for the prediction profiles - ordering by decision values
# heatmap(predProf, Rowv="decision", main="Prediction Profiles", labels=yCC,
# RowSideColors=c("blue", "red"), cexRow=0.15, cexCol=0.3)
# 
# ## plot heatmap for the prediction profiles - with hierarchical clustering
# heatmap(predProf, Rowv=TRUE, main="Prediction Profiles", labels=yCC,
# RowSideColors=c("blue", "red"), cexRow=0.15, cexCol=0.3)
# ## End(Not run)