TANOVA (version 1.0.0)

tanova: Classification of genes by time course analysis of variance(TANOVA)

Description

The method is useful to capture gene specific response during a time course and their dependency on multiple experimental factors. It is based on non-parametric ANOVA technique. It's applicable to both longitudinal and cross-sectional data. gene.classifier is the major function users should call. gene.classifier1/gene.classifier2/gene.classifier3 are internal functions called by gene.classifier for specific analysis situation.

Usage

tanova(data, f1, f2, tp, B=100, FDR=0.05,robustify=FALSE, equal.size=FALSE, qt=-1, longitudinal=TRUE, test.type=0,eb=FALSE,df=0)
gene.classifier1(data, f1, f2, B=100,FDR=0.05, robustify=FALSE, equal.size=FALSE, eb=FALSE,qt=-1)
gene.classifier2(data, f1, f2, B=100,FDR=0.05, robustify=FALSE, equal.size=FALSE, time.course, qt=-1,eb=FALSE,df=1)
gene.classifier3(data, f1, f2, tp, B=100,FDR=0.05, qt=-1,robustify=FALSE,eb=FALSE)

Arguments

data

data matrix (gene * array). Each row is a gene. Each column is an array. If data are longitudinal (for example, time course measurements from patients), arrays from same experimental units (e.g. patient) should be adjacent to each other.

f1

a vector with length equal to the number of arrays. Each entry indicates the level of the first factor for corresponding array. The values of f1 should be 1,2,3,...

f2

a vector with length equal to the number of arrays. Each entry indicates the level of the second factor for the corresponding array. The values of f2 should be 1,2,3,... If the experimental has only one factor, let f2=0.

tp

a vector with length equal to the number of arrays. Each entry indicates the time point for the corresponding array. tp takes values 1,2,3 .... For non-time course data, let tp=0.

B

the number of bootstrap resampling. Default is 100. Large B lead to more accurate inference, but need more running time.

FDR

false discovery rate (FDR) for each test. Default is 0.05.

robustify

a logical indicator of whether a robust test statistic should be used. Default is FALSE.

equal.size

a logical indicator of whether the number of replicates under each biological condition is equal. Default is FALSE.

qt

a vector of quantiles for which FDR is estimated by averaging over B times of number of false genes. If qt=0.25 (0.5, 0.75, 0.9), FDR is estimated by 25% (50%, 75%, 90%) quantiles of false genes over B times.Default is -1, which means all genes are used.

longitudinal

a logical indicator of whether the data is longitudinal. The default is TRUE

eb

a logical indicator of whether Empirical Bayesian method should be used in the estimation of significance

test.type

an indicator of TANOVA test type. 0: classifies genes into gene sets C1,C2, C3,C4 and C5 (constant genes). 1: test for interaction effect. 2: one-way NANOVA test. 3: test main effect f1. 4: test main effect f2.

df

degree of freedom

time.course

the number of time points we sampled

Value

Depends on the test.type user specified, the output will be different. If test.type=0, the output is a list object containing four classes of genes that response to factors differently determined by tanova. They are described below:

C1, C1.delta, C1.pvalue, a1

the gene index, test statistic, p-value and projection vector of genes that show significant interaction effect of two treatment factors (f1*f2)

C2, C2.delta, C2.pvalue, a2

the gene index, test statistic, p-value and projection vector of genes that show significant additive effect of two treatment factors (f1+f2)

C3, C3.delta, C3.pvalue, a3

the gene index, test statistic, p-value and projection vector of genes that show significant effect of treatment factor #1 (f1)

C4, C4.delta, C4.pvalue, a4

the gene index, test statistic, p-value and projection vector of genes that show significant effect of treatment factor #2 (f2)

If test.type=1,2,3,4. Only one of the above class will be returned. The items in the output list are described as follows:

genes

A numeric vector indicating the positions in which the genes are called significant for the test

pvalue

a numeric vector of the corresponding p-value of TANOVA.

delta

a numeric vector of summary statistic for non-parametric ANOVA

a

a matrix containing projection direction (gene by time point)

dir

a list object returned by proj.dir

obj

a list object returned by NANOVA.test

See Also

NANOVA.test

Examples

	##f1=rep(1:2, each=8)
	##f2=rep(c(1,2,1,2), each=4)
	##tp=rep(1:4, 4)
	##data=matrix(rnorm(16*1000), nrow=1000, ncol=16)
	##result=gene.classifier(data,f1,f2,tp)