Download Gene Expression Analysis of Time Course Experiment - Lecture Slides | STAT 5570 and more Study notes Statistics in PDF only on Docsity! Gene expression analysis of time course experiments Reed Gann Statistical bioinformatics 6570 Spring ‘09 References • Conesa, M. Nueda, A. Ferrer, and M. Tal´on. masigpro: a method to identify significant differential expression profiles in time course microarray experiments., 2005. • Tusher, V., Tibshirani, R. and Chu, G. (2001): Significance analysis of microarrays applied to the ionizing radiation response" PNAS 2001 98: 5116-5121 Replicates • Not discussed previously • Allow us to get a better estimate of variance • Define Biological vs technical replicates • Rep plot; additional quality check required Rep Plot
library (affy)
data <- ReadAffy(celfile.path="~/Documents/idisk
local/Spring '09/Statistical Bioinformatics/TIMECOURSE"
Data <- rma(data)
exprs.Data <- exprs(Data)
dim(exprs.Data)
plot(exprs.Data[,1],exprs.Data[,2], pch=16, cex=.8,
xlab='Trt0tp30Rep1', ylab='Trt0tp30Rep2' )
abline(0,1)
Rep Plot
RG13-TRT4-TP120-REP2.CEL
2 4 6 8 10 2 14 RG34-4-120-1a.CEL
maSigPro
library (affy)
data <- ReadAffy(celfile.path="~/Documents/idisk
local/Spring '09/Statistical Bioinformatics/TIMECOURSE")
#prepare data#
Data <- rma(data)
exprs.Data <- exprs(Data)
T.cell <- c(rep(0,6),rep(1,6))
# O=control, 1=treatment
gn <- rownames(exprs.Data)
maSigPro.data <- list(x=exprs.Data, y=(T.cell),
geneid=gn, genenames=gn, logged2=TRUE)
colnames(exprs.Data) <- paste("Array", c(1:12), sep = "")
#create experimental design#
Time <- rep(c(rep(c(1:3), each = 2)),
Replicates <- c(rep(c(1:6), each = 2))
Control <- c(rep(1, 6), rep(0, 6))
Treatl <- c(rep(0, 6), rep(1, 6))
edesign <- cbind(Time, Replicates, Control, Treatl)
2)
#### RUN maSigPro
library(maSigPro)
tc.test <- maSigPro (exprs.Data, edesign, degree = 1, vars
= “groups", main = "Test")
tc.test$summary # shows significant genes by experimental
groups
maSigPro • NO significant genes?!? • This is a result of the model maSigPro model • “maSigPro follows a two step regression strategy to find genes with significant temporal expression changes and significant differences between experimental groups”. • Does not make any mention of repeated measures or accounting for time-treatment interactions • Major issue-must characterize highest order interactions if significance is found doing a post hoc analysis SAM • We get significant genes at an FDR of 5% • At delta 1.68 get about 30 significant genes • Model is more appropriate SAM • Recall from our previous visit to SAM • “relative difference” function • s(i) is “gene-specific scatter” and is defined as the standard deviation of repeated expression measurements SAM vs. maSigPro • SAM model accounts for repeated measurements that are a result of a time course experimental design-maSigPro is simply a two step regression method applied to each of the experimental factors • Costly error-Significant genes vs. no significant genes • Post-hoc analysis proves that the time/treatment interaction is significant (not shown)