SortModes: Sorting of ICA Modes

In mlica2: Independent Component Analysis using Maximum Likelihood

Description

Sorts inferred ICA modes using two criteria: Relative data power or the Liebermeister criterion, which is based on a measure that is a weighted linear combination of non-gaussianity and data variance measures.

Usage

SortModes(a.best,c.val = 0.25)

Arguments

a.best	The output object of 'mlica'.
c.val	A parameter to control the relative weight of the two measures when using the Liebermeister criterion. Should be between 0 (pure data variance measure) and 1 (pure non-gaussianity).

Value

A list with the components:

a.best: The output of 'mlica'.

rdp: The relative data power values obtained for each independent component.

lbm: The Liebermeister contrast value for each component.

Author(s)

Andrew Teschendorff [e-mail]

References

Hyvaerinen A., Karhunen J., and Oja E.: Independent Component Analysis, John Wiley and Sons, New York, (2001).

Kreil D. and MacKay D. (2003): Reproducibility Assessment of Independent Component Analysis of Expression Ratios from DNA microarrays, Comparative and Functional Genomics *4* (3),300-317.

Liebermeister W. (2002): Linear Modes of gene expression determined by independent component analysis, Bioinformatics *18*, no.1, 51-60.

Examples

#This function is currently defined as
function(a.best,c.val=0.25){
  
	ncp <- ncol(a.best$S);
	Ng <- nrow(a.best$S);

	#SORTING CRITERION

	# A) Computation of relative data power. Store values in vector of size H=ncp. Could use different criterion here.
	# need squared entries
	 Ssq <- a.best$S * a.best$S ; 
	 Asq <- a.best$A * a.best$A ;
	 Xsq <- a.best$X * a.best$X ;
	 rdp <- rep(0, times=ncp);
	 for ( k in 1:ncp ) {
	  rdp[k]<- sum(Ssq[,k])*sum(Asq[k,])/sum(Xsq) ;
	 }
	 rdp.s <- sort(rdp, na.last=NA,decreasing=TRUE, index.return=TRUE);

	# B) sorting with mixture of contrast and data variance (Liebermeister)
	 JG <- rep(0, times=ncp);
	 JA <- rep(0, times=ncp);
	 # generate values from std. normal distribution
	 nu <- rnorm(10000,0,1);
	 G0 <- mean(log(cosh(nu)));
	 
	 for ( k in 1:ncp ){
	   # compute contrast for mode using logcosh
	   G1 <- mean(log(cosh(a.best$S[,k])));
	   JG[k] <- abs(G1-G0);
	   JA[k] <- sum(Asq[k,]);
	 }   
	   sumJG <- sum(JG) ; sumJA <- sum(JA) ;
	   J <- JG*(c.val/sumJG)+ JA*(1-c.val)/sumJA ;
	   J.s <- sort(J, na.last=NA,decreasing=TRUE, index.return=TRUE);

	   return(list(a.best=a.best,rdp=rdp.s,lbm=J.s));

	} 

mlica2 documentation built on May 1, 2019, 10:45 p.m.