[BioC] Visualizing pre- and post-normalization for single-colour arrays

[mrjmorri at ucalgary.ca]

Hello,

My name is Matthew Morris, I'm a second year Master's student at the
 University of Calgary, and I am fairly new to the world of microarrays!
 My project involves single-colour arrays, for which helpful documents seem
 to be rather limited.

As such, I have some questions.  I'll ask the questions first, and then
 show you the code.  If you can only answer one of the three, some info is
 better than none!

1. How do I visualize my data pre- and post-normalization?  What will I be
 looking for?

2. I am using code from someone else to flag nonuniform and feature
 population outliers.  It certainly alters my results, but I'm not sure how
 to check if it is working correctly.

3. How can I incorporate things like sex and length into my model? (to
 clarify, I am looking at four populations of fish called Cran, Hog, OL and
 LCM respectively, raised at 7 or 23 degrees, and I would like to eliminate
 the effects of sex, tank and length)


Thank you very much,
Matthew Morris

Code is as follows:

library(limma)

#read Targets file (make sure to set directory first through File)
 targets<-readTargets("targets.txt")

#checks that file was read correctly
targets$FileName

#weight OL

wtAgilent.mRGOLFilter <- function(qta)
{mapply(min,1-qta[,"gIsFeatNonUnifOL"],1-qta[,"gIsFeatNonUnifOL"],1-qta[,"gIsFeatPopnOL"],1-qta[,"gIsFeatPopnOL"])}

#read data from array ouput files into E
E<-read.maimages(targets$FileName,
source="agilent.median",path="actualall", wt.fun=wtAgilent.mRGOLFilter,
 columns=list(E="gProcessedSignal"),
other.columns=list(saturated="gIsSaturated",
nonuniform="gIsFeatNonUnifOL", popnoutlier="gIsFeatPopnOL",
flag="IsManualFlag", wellabovebg="gIsWellAboveBG"))

#to see that everything looks fine
E

#normalizing between arrays:
normalize<-normalizeBetweenArrays(E, method="quantile")


#remove control spots
dat1<-normalize[normalize$genes$ControlType==0,]

#average values for identical probes within an array
Eavg<-avereps(dat1, ID=dat1$genes$ProbeName)


#analyze factorial design: first identify the factors in template
 TS <- paste(targets$Population, targets$Temperature, sep=".")

#check to see it worked
TS

#set up design
TS <- factor(TS, levels=c("Hog.7","Cran.7","LCM.7","OL.7", "Hog.23",
 "Cran.23", "OL.23", "LCM.23", "Hog.15"))
design <- model.matrix(~0+TS)
colnames(design) <- levels(TS)
fit <- lmFit(Eavg, design)
cont.matrix <- makeContrasts(Hog.7vs23=Hog.23-Hog.7,
Cran.7vs23=Cran.23-Cran.7, LCM.7vs23=LCM.23-LCM.7, OL.7vs23=OL.23-OL.7,
 levels=design)
fit2 <- contrasts.fit(fit, cont.matrix)
fit2 <- eBayes(fit2)

#check results

topTable(fit2, coef=1, adjust="BH")
results <- decideTests(fit2)
vennCounts(results)
vennDiagram(results)

write.table (fit2, file="fit2.txt")