z score transformation from signal intensities of raw microarray data

Question

[iphoenix2100]

Dear all,

I am making this slow transition from GWAS findings to expression datasets. We are using an online repository of microarray data from different donors (arrayed for different regions of the brain). And basically we are using a set of genes from the initial GWAS findings and see how it is affecting the expression pattern in different regions of the brain.

I came across this method of z-score transformation of the values in RAW microarray data to make sense of whether the gene is up or down regulated i.e (z-score values were calculated by subtracting the total average intensity of the genes in the microarray from the intensity of each individual gene within a single experiment and dividing the result by the standard deviation (SD) of the measure of all the intensities)

Of the multiple headers in the raw data, I believe (the formula is referring to gProcessedSigError for SD; gMeanSignal for signal intensity genes in the array)

My question - is these two columns are enough to calculate the z-score or am I missing something. Sorry for a more specific and a basic question. Any suggestions will be of great help

Thanks

ps.

Some other columns may be relevance

gSurrogateUsed
gIsFound
gProcessedSignal
gProcessedSigError
gNumPixOLHi
gNumPixOLLo
gNumPix
gMeanSignal
gMedianSignal
gPixSDev
gPixNormIQR
gBGNumPix
gBGMeanSignal
gBGMedianSignal
gBGPixSDev
gBGPixNormIQR
gNumSatPix

Answer 1

[Manvendra Singh]

##### First you need to process your data, if its Illumina microarray then
library(lumi) #### load the library
dati <- lumiR("non-normalized_data.txt", columnNameGrepPattern = list(exprs = "AVG_signal",
se.exprs = "BEAD_STERR", beadNum = NA, detection = "Detection Pval"))
dati.T = lumiT(dati, method = "log2") # transform data to log2
dati.qnorm = lumiN(dati.T, method = "quantile",verbose = FALSE)  ###### quantile normalization of data
data <- exprs(dati.qnorm) ######## get the expression table of your data and now check if

colnames(data) ###### is your sample names and

row.names(data)    ######is your gene names

then z-score calculation is

scaled.data <- scale(data)   ######## scaled.data is have z-score of each gene across the samples

hth

[iphoenix2100]

Thanks Mr. Singh,,

It seems the data in repository is all normalized microarray expression values or that is what the readme file of the data says ..

what options do I have in that case - can I calculate the zscore by the methods that i have mentioned in my query or not! thanks again for your suggestions