fitPrior: Fit the expression value profiles with a mixture of normal components

Description

firPrior performs a clustering of expression values for each gene profile using the mclust function. This results to a mixture of normal distribution components (from 1 to 3 components) fitting the expression values.

Usage

fitPrior(expressionMatrix, param.detection = NULL, lambda = 1, 
beta = 6, evaluation.lambda.beta = FALSE)

Arguments

expressionMatrix

the expression value matrix, genes*conditions

param.detection

the matrix of parameters as obtained by getDefaultParameter or createParamterMatrix. It must contain positive values for "lambda" and "beta". If NULL, the function getDefaultParameter will be used

lambda

positive value, it influences the choice of models by affecting the selection of one, two or three normal distributions, thus introducing some weight on the effect of number of parameters to be defined. The default is 1, the model uses the BIC value taking into account the log-likelihood value

beta

positive value, it influences the prior applied during the determination of the variance of the normal distributions. It is important for this fitting since it allows the model to capture isolated outliers. The default value is 6

evaluation.lambda.beta

if TRUE, an extra attribute will be return indicating for how many gene the lambda and beta parameters change the number of normal component chosen to fit the expression values

Value

fit1: list of the gene as first attributes, for each gene a list of three attributes:
G: number of normal components fitting the data
NorMixParam: the parameters of each normal component: proportion, mean and standard deviation for the gene
classification: the normal component id in which the expression values of the gene are attributed

Details

If evaluation.lambda.beta is TRUE an additional attribute G.lambda.beta.effect is returned. It is a matrix presenting the number of time the values of G (number of normal components for a particular gene) has changed between lambda=0 and the lambda.1 value and between beta=0 and the beta.1 value.

Examples

Run this code

library(SpeCond)
data(expressionSpeCondExample)
##Perform the SpeCond analysis step by step
param.detection=getDefaultParameter()
param.detection

fit1=fitPrior(expressionSpeCondExample, param.detection=param.detection)

##then use getSpecificOutliersStep1(), fitNoPriorWithExclusion() and 
## getSpecificResult()