bootsPLS (version 1.1.2)

bootsPLS-package: bootsPLS

Description

bootsPLS

Details

The DESCRIPTION file: bootsPLS bootsPLS The package implements the methodology described in Rohart et al. (2016) for identifying genes that differentiate Mesemchymal Stromal Cells from other cell types. The method is applicable to any classification problem with more than 2 classes. It relies on bootstrap subsamplings of sPLS-DA and provides tools to select the most stable variables (defined as the ones consistently selected over the bootstrap subsamplings) and to predict the class of test samples.

Three major functions: *bootsPLS performs a tune.splsda of the mixOmics package on several random subsamplings. It records the selected variables on each replication. The compile.bootsPLS.object function is to be used if several calls to bootsPLS are made. *fit.model fits a constraint spls (see spls.hybrid) on the most stable variables. The variables can be automatically selected by setting auto.tune=TRUE. *prediction predicts the class of test samples and compute Confidence Interval (CI) of the prediction values.

References

Rohart et al. (2016). A Molecular Classification of Human Mesenchymal Stromal Cells. PeerJ, DOI 10.7717/peerj.1845 Le Cao, K.-A., Boitard, S. and Besse, P. (2011). Sparse PLS Discriminant Analysis: biologically relevant feature selection and graphical displays for multiclass problems. BMC Bioinformatics 12:253.

Examples

# NOT RUN {
data(MSC)
X=MSC$X
Y=MSC$Y


boot1=bootsPLS(X=X,Y=Y,ncomp=3,many=5,kCV=5)
boot2=bootsPLS(X=X,Y=Y,ncomp=3,many=5,kCV=5,showProgress=FALSE)
boot3=bootsPLS(X=X,Y=Y,ncomp=3,many=5,kCV=5,showProgress=FALSE, cpus=2)

# construct a list of bootsPLS object
bootsPLS.object=list(boot1,boot2,boot3)

# compile the outputs in one bootsPLS object
boot=compile.bootsPLS.object(bootsPLS.object)

# fit the model
fit=fit.model(boot,auto.tune=TRUE) #tuning number of components and variables
fit=fit.model(boot,ncomp=2) #tuning number of variables on the 2components
plotIndiv(fit,ind.names=FALSE, legend=TRUE)

# prediction and Confidence Interval, here we use the same dataset but it should be external data
pred=prediction(fit,X.test=X,CI=TRUE)

head(pred$Y.hat.test[,,"comp.1"])
lapply(pred$out.CI$CI$'comp.1',head)

# }