svpls: Function for identfying the optimal ANCOVA model and detecting the genes that are truly differentially expressed between the two types of samples.

Description

This function calls fitModel repeatedly to fit a series of ANCOVA models along with the standard ANOVA model, to the log transformed gene expression data. The model with the minimum AIC is selected as the optimal one and its corresponding estimated effects are then used to perform a multiple testing of differential expression, over all the genes, using the Benjamini-Hochberg correction.

Usage

svpls(k1, k2, Y, pmax = 3, fdr = 0.05)

Arguments

Number of subjects/samples under variety 1.

Number of subjects/samples under variety 2.

The log transformed gene expression data, with genes along the rows and subjects/samples along the columns.

pmax

Maximum number of surrogate variables to be incorporated in the ANCOVA model (means pmax ANCOVA models are fitted to the data). By default, it is taken as 3.

fdr

The specified False Discovery Rate (FDR) for multiple testing of differential expression, using the Benjamini-Hochberg correction. By Default it is taken as 0.05.

Value

svpls returns an object of class "svpls". The object contains a list of the following components:
opt.modelThe optimal model. 1 denotes the standard ANOVA model.
PLS.impPlS imputed estimate of the hidden expression heterogeneity, evaluated from the optimal model (applicable only when opt.model>1).
Y.corrCorrected gene expression matrix after adjusting for the hidden effects (applicable only when opt.model>1).
pvaluesp-values from the tests with the effects estimated from the standard ANOVA model (returned only when opt.model=1.
pvalues.adjAdjusted p-values after correcting for the hidden effects (applicable only when opt.model>1).
genesGenes that are deemed to be differentially expressed from the multiple hypotheses testing with effects estimated from the optimal model.
AIC.optAIC value for the optimal model.

References

Hirotsugu, A. (1980) Likelihood and the Bayes Procedure. The Institute of Statistical Mathematics, Tokyo., Benjamini, Y and Hochberg, Y (1995) Controlling the false discovery rate : a practical and powerful approach to multiple testing. Journal of the Royal Statistical Society.

Examples

Run this code

## Loading the first dataset
data(batch.dat)

## Fitting the optimal ANCOVA model to the data gives:
fit <- svpls(10,10,batch.dat,pmax = 5)

## The optimal ANCOVA model, its AIC value and the positive genes detected from it are given by:
fit$opt.model 

fit$AIC.opt

fit$genes

## The corrected gene expression matrix obtained after removing the effects of the hidden variability is given by:
Y.corrected <- fit$Y.corr