SMVar.unpaired {SMVar}

Structural model for variances with unpaired data

Description

Function to detect differentially expressed genes when data are unpaired

Usage

SMVar.unpaired(geneNumbers, listcond, fileexport = NULL,
               minrep = 2, method = "BH", threshold = 0.05)

Arguments

geneNumbers

Vector with gene names or dataframe which contains all information about spots on the chip

listcond

list of the different conditions to be compared

fileexport

file to export the list of differentially expressed genes

minrep

minimum number of replicates to take a gene into account, minrep must be higher than 2

method

method of multiple tests adjustment for p.values

threshold

threshold of False Discovery Rate

Details

This function implements the structural model for variances described in (Jaffrezic et al., 2007). Data must be normalized before calling the function. Matrix geneNumbers must have one of the following formats: "matrix","data.frame","vector","character","numeric","integer".

Value

Only the number of differentially expressed genes is printed. If asked, the file giving the list of differentially expressed genes is created.

If the user creates an object when calling the function (for example "Stat=SMVar.paired(...)") then Stat contains the information for all genes, is sorted by ascending p-values and

Stat$TestStat

gives the test statistics as described in the paper

Stat$StudentPValue

gives the raw p-values

Stat$DegOfFreedom

gives the number of degrees of freedom for the Student distribution for the test statistics

Stat$Cond1

gives the first condition considered in the log-ratio

Stat$Cond2

gives the second condition considered in the log-ratio

Stat$LogRatio

gives the logratios (listcond[[Cond2]]-listcond[[Cond1]])

Stat$AdjPValue

gives the adjusted p-values

Note

If the first column of the file geneNumbers contains identical names for two different spots, these two spots are only counted once if they are both differentially expressed. By default, the correction for multiple testing is Benjamini Hochberg with a threshold of False Discovery Rate (FDR) of 5%. The FDR threshold can be changed, and it is also possible to choose the multiple test correction method ("holm", "hochberg", "hommel", "bonferroni", "BH", "BY", "fdr", "none"). To see the references for these methods, use the R-help ?p.adjust.

Author(s)

Guillemette Marot with contributions from Anne de la Foye

References

F. Jaffrezic, Marot, G., Degrelle, S., Hue, I. and Foulley, J. L. (2007) A structural mixed model for variances in differential gene expression studies. Genetical Research (89) 19:25

Examples

library(SMVar)
data(ApoAIdata)
attach(ApoAIdata)
SMVar.unpaired(ApoAIGeneId,list(ApoAICond1,ApoAICond2))