R package for estimating optimal number of biomarkers at a given error tolerance level for various classification rules

Description

Using interactive control panel (rpanel) and 3D real-time rendering system (rgl) , this package provides a user friendly GUI for estimating the minimum number of biomarkers (variables) needed to achieve a given level of accuracy for two-group classification problems based on microarray data.

Details

The function optimiseBiomarker is a user friendly GUI for interrogating the database of leave-one-out cross-validation errors, errorDbase, to estimate optimal number of biomarkers for microarray based classifications. The database is built on the basis of simulated data using the classificationError function. The function simData is used for simulating microarray data for various combinations of factors such as the number of biomarkers, training set size, biological variation, experimental variation, fold change, replication, and correlation.

Author(s)

Mizanur Khondoker, Till Bachmann, Peter Ghazal

Maintainer: Mizanur Khondoker mizanur.khondoker@gmail.com.

References

Khondoker, M. R., Till T. Bachmann, T. T., Mewissen, M., Dickinson, P. et al.(2010). Multi-factorial analysis of class prediction error: estimating optimal number of biomarkers for various classification rules. Journal of Bioinformatics and Computational Biology, 8, 945-965.

Breiman, L. (2001). Random Forests, Machine Learning 45(1), 5–32.

Chang, Chih-Chung and Lin, Chih-Jen: LIBSVM: a library for Support Vector Machines, https://www.csie.ntu.edu.tw/~cjlin/libsvm/.

Ripley, B. D. (1996). Pattern Recognition and Neural Networks.Cambridge: Cambridge University Press.

Efron, B. and Tibshirani, R. (1997). Improvements on Cross-Validation: The .632+ Bootstrap Estimator. Journal of the American Statistical Association 92(438), 548–560.

Bowman, A., Crawford, E., Alexander, G. and Bowman, R. W. (2007). rpanel: Simple interactive controls for R functions using the tcltk package. Journal of Statistical Software 17(9).

See Also

simData classificationError optimiseBiomarker

Examples

if(interactive()){
data(errorDbase)
optimiseBiomarker(error=errorDbase)
}

Estimation of misclassification errors (generalisation errors) based on statistical and various machine learning methods

Description

Estimates misclassification errors (generalisation errors), sensitivity and specificity using cross-validation, bootstrap and 632plus bias corrected bootstrap methods based on Random Forest, Support Vector Machines, Linear Discriminant Analysis and k-Nearest Neighbour methods.

Usage

## S3 method for class 'data.frame'
classificationError(
          formula,
          data, 
          method=c("RF","SVM","LDA","KNN"), 
          errorType = c("cv", "boot", "six32plus"),
	  senSpec=TRUE,
          negLevLowest=TRUE,
	  na.action=na.omit, 
          control=control.errorest(k=NROW(na.action(data)),nboot=100),
          ...)

Arguments

Details

In the current version of the package, estimation of sensitivity and specificity is limited to cross-validation estimator only. For LDA sample size must be greater than the number of explanatory variables to avoid singularity. The function classificationError does not check if this is satisfied, but the underlying function lda produces warnings if this condition is violated.

Value

Returns an object of class classificationError with components

Author(s)

Mizanur Khondoker, Till Bachmann, Peter Ghazal
Maintainer: Mizanur Khondoker mizanur.khondoker@gmail.com.

References

Khondoker, M. R., Till T. Bachmann, T. T., Mewissen, M., Dickinson, P. et al.(2010). Multi-factorial analysis of class prediction error: estimating optimal number of biomarkers for various classification rules. Journal of Bioinformatics and Computational Biology, 8, 945-965.

Breiman, L. (2001). Random Forests, Machine Learning 45(1), 5–32.

Chang, Chih-Chung and Lin, Chih-Jen: LIBSVM: a library for Support Vector Machines, https://www.csie.ntu.edu.tw/~cjlin/libsvm/.

Ripley, B. D. (1996). Pattern Recognition and Neural Networks.Cambridge: Cambridge University Press.

Efron, B. and Tibshirani, R. (1997). Improvements on Cross-Validation: The .632+ Bootstrap Estimator. Journal of the American Statistical Association 92(438), 548–560.

See Also

simData

Examples

## Not run: 
mydata<-simData(nTrain=30,nBiom=3)$data
classificationError(formula=class~., data=mydata)

## End(Not run)

Database of leave-one-out cross validation errors for various combinations of data characteristics

Description

This is a 7-dimensional array (database) of leave-one-out cross validation errors for Random Forest, Support Vector Machines, Linear Discriminant Analysis and k-Nearest Neighbour classifiers. The database is the basis for estimating the optimal number of biomarkers at a given error tolerance level using optimiseBiomarker function. See Details for more information.

Usage

data(errorDbase)

Format

7-dimensional numeric array.

Details

The following table gives the dimension names, lengths and values/levels of the data object errorDbase.

We have a plan to expand the database to a 8-dimensional one by adding another dimension to store error rates at different level of correlation between biomarkers. Length of each dimension will also be increased leading to a bigger database with a wider coverage of the parameter space. Current version of the database contain error rates for independent (correlation = 0) biomarkers only. Also, it does not contain error rates for Linear Discriminant Analysis, which we plan to implement in the next release of the package. With the current version of the database, optimal number of biomarkers can be estimated using the optimiseBiomarker function for any intermediate values of the factors represented by the dimensions of the database.

Author(s)

Mizanur Khondoker, Till Bachmann, Peter Ghazal
Maintainer: Mizanur Khondoker mizanur.khondoker@gmail.com.

References

Khondoker, M. R., Till T. Bachmann, T. T., Mewissen, M., Dickinson, P. et al.(2010). Multi-factorial analysis of class prediction error: estimating optimal number of biomarkers for various classification rules. Journal of Bioinformatics and Computational Biology, 8, 945-965.

See Also

optimiseBiomarker

Internal functions for optBiomarker package.

Description

Internal functions for optBiomarker package.

Usage

tnormAvg(a1=-Inf, a2=Inf,mu=0,sigma=1)
avg2Lower(truncAvg=log2(2), mu=0,sigma=1)
objFun(para, mu=0,sigma=1,truncAvg=log2(2))
plot3dFun(panel,...)
dimSelect(x, val)
yapprox(y, x, xout, method="linear", n=50,
          yleft, yright, rule = 1, f = 0, ties = mean)
matapprox(mat, rx, cx, nr=100,nc=20,rout, cout)

classpredict.lda(object, newdata)
classpredict.knn(object, newdata) 

Details

These functions are not to be called by the user.

Estimates optimal number of biomarkers at a given error tolerance level for various classification rules

Description

Using interactive control panel (see rpanel) and 3D real-time rendering system (rgl), this package provides a user friendly GUI for estimating the minimum number of biomarkers (variables) needed to achieve a given level of accuracy for two-group classification problems based on microarray data.

Usage

optimiseBiomarker (error,
                   errorTol = 0.05,
                   method = "RF", nTrain = 100, 
                   sdB = 1.5,
                   sdW = 1,
                   foldAvg = 2.88,
                   nRep = 3) 

Arguments

Details

The function optimiseBiomarker is a user friendly GUI for interrogating the database of leave-one-out cross-validation errors, errorDbase, to estimate optimal number of biomarkers for microarray based classifications. The database is built on the basis of simulated data using the classificationError function. The function simData is used for simulating microarray data for various combinations of factors such as the number of biomarkers, training set size, biological variation, experimental variation, fold change, replication, and correlation.

Author(s)

Mizanur Khondoker, Till Bachmann, Peter Ghazal

Maintainer: Mizanur Khondoker mizanur.khondoker@gmail.com.

References

Khondoker, M. R., Till T. Bachmann, T. T., Mewissen, M., Dickinson, P. et al.(2010). Multi-factorial analysis of class prediction error: estimating optimal number of biomarkers for various classification rules. Journal of Bioinformatics and Computational Biology, 8, 945-965.

Breiman, L. (2001). Random Forests, Machine Learning 45(1), 5–32.

Chang, Chih-Chung and Lin, Chih-Jen: LIBSVM: a library for Support Vector Machines, https://www.csie.ntu.edu.tw/~cjlin/libsvm/.

Ripley, B. D. (1996). Pattern Recognition and Neural Networks.Cambridge: Cambridge University Press.

Efron, B. and Tibshirani, R. (1997). Improvements on Cross-Validation: The .632+ Bootstrap Estimator. Journal of the American Statistical Association 92(438), 548–560.

Bowman, A., Crawford, E., Alexander, G. and Bowman, R. W. (2007). rpanel: Simple interactive controls for R functions using the tcltk package. Journal of Statistical Software 17(9).

See Also

simData classificationError

Examples

if(interactive()){
data(errorDbase)
optimiseBiomarker(error=errorDbase)
}

A set of 54359 median gene expressions in log (base 2) scale

Description

This data set contains a set of 54359 log base 2 gene expression values from a neonatal whole blood gene expression study described in Smith et al. (2007). The data represent the median of 28 microarrays corresponding to 28 control (healthy) patients of the neonatal study. This data set is used as a base expressions set for simulating biomarker data using simData function of the optBiomarker package.

Usage

data(realBiomarker)

Format

A vector of 54359 gene expressions in log (base 2) scale.

References

Smith, C. L., Dickinson, P., Forster, T., Khondoker, M. R., Craigon, M., Ross, A., Storm, P., Burgess, S., Lacaze, P., Stenson, B. J.and Ghazal, P. (2007). Quantitative assessment of whole blood RNA as a potential biomarker for infectious disease. Analyst 132, 1200–1209.

Khondoker, M. R., Till T. Bachmann, T. T., Mewissen, M., Dickinson, P. et al.(2010). Multi-factorial analysis of class prediction error: estimating optimal number of biomarkers for various classification rules. Journal of Bioinformatics and Computational Biology, 8, 945-965.

Simulation of microarray data

Description

The function simulates microarray data for two-group comparison with user supplied parameters such as number of biomarkers (genes or proteins), sample size, biological and experimental (technical) variation, replication, differential expression, and correlation between biomarkers.

Usage

simData(nTrain=100,
        nGr1=floor(nTrain/2),
        nBiom=50,nRep=3,
        sdW=1.0,
        sdB=1.0,
        rhoMax=NULL, rhoMin=NULL, nBlock=NULL,bsMin=3, bSizes=NULL, gamma=NULL,
        sigma=0.1,diffExpr=TRUE,
        foldMin=2,
        orderBiom=TRUE,
        baseExpr=NULL)

Arguments

Details

Differential expressions are introduced by adding z\delta to the data of group 2 where \delta values are generated from a truncated normal distribution and z is randomly selected from (-1,1) to characterise up- or down-regulation.

Assuming that Y ~is~ N(\mu, \sigma^2), and A=[a_1,a_2], a subset of -Inf <y < Inf, the conditional distribution of Y given A is called truncated normal distribution:

f(y, \mu, \sigma)= (1/\sigma) \phi((y-\mu)/\sigma) / (\Phi((a2-\mu)/\sigma) - \Phi((a_1-\mu)/\sigma))

for a_1 <= y <= a_2, and 0 otherwise,

where \mu is the mean of the original Normal distribution before truncation, \sigma is the corresponding standard deviation,a_2 is the upper truncation point, a_1 is the lower truncation point, \phi(x) is the density of the standard normal distribution, and \Phi(x) is the distribution function of the standard normal distribution. For simData function, we consider a_1=log_2(\code{foldMin}) and a_2=Inf. This ensures that the biomarkers are differentially expressed by a fold change of foldMin or more.

Value

A dataframe of dimension nTrain by nBiom+1. The first column is a factor (class) representing the group memberships of the samples.

Author(s)

Mizanur Khondoker, Till Bachmann, Peter Ghazal
Maintainer: Mizanur Khondoker mizanur.khondoker@gmail.com.

References

Khondoker, M. R., Till T. Bachmann, T. T., Mewissen, M., Dickinson, P. et al.(2010). Multi-factorial analysis of class prediction error: estimating optimal number of biomarkers for various classification rules. Journal of Bioinformatics and Computational Biology, 8, 945-965.

See Also

classificationError

Examples

simData(nTrain=10,nBiom=3)