Analyzing normalized microarray data

Question

[klarcha]

Hi everyone,

I received a set of microarray data that has already been processed and normalized by the microarray core facility (essentially a matrix with probe IDs, expression values for each sample and a note if the gene is 'expressed over background').

The experimental setup is as follows: we have 3 groups, 6 subjects per group, and each subject was sampled 3 times (T1, T2 and T3) over a course of a month (with intermediate interventions) - so the data is paired. The comparison I want to make is T2vT1, T3vT1 and T3vT2 within each group (to see how the gene expression changes upon intervention), and I want to calculate Log2FC and p-values to produce volcano plots.

I assume the best R package for this type of analysis is limma - however I cannot find a way how to import normalized data into it, and where to start the analysis.

Any help or thoughts would be greatly appreciated!

Thanks in advance.

[ATpoint]

What is the current format of the data?

[klarcha]

A matrix with normalized expression values, encoded as a .csv file. The 1st column contains Probe IDs and the 1st row sample IDs.

Answer 1

[Benn]

You need to import the matrix into R and make an ExpressionSet of it (https://www.rdocumentation.org/packages/Biobase/versions/2.32.0/topics/ExpressionSet). You can keep it as minimal as you want, or add stuff like pheno data and annotation. With the ExpressionSet you can continue with limma.

Answer 2

[Gordon Smyth]

Generally speaking, it is better to get the raw microarray data from your core facility and allow limma to read and normalize the raw files. Then you know what is happening.

If you must work with the already processed csv file, then just read it into R using read.csv. You will have to wrangle it into a data matrix containing normalized log-intensity values and the Probe IDs as row.names. Once you have a matrix, limma can work on that directly.