Normalized intensities from Beadchip expression arrays to find gene coexpressions?

Question

[tigeradab]

I'm using publicly available microarray data for finding correlations in gene expression. I am a novice in microarray data analysis and usage. As I understand, microarray data is better used for differential analysis and cannot be used as a surrogate for absolute expression. But it seems logical for me, that changes in signal intensity across conditions can be used for finding expression correlations between genes. Am I wrong in assuming this?

In order to do this, I also require unique expression vectors from each gene. But as I understand about Illumina Beadchip data, there are multiple copies of a probe associated to a gene (probably signifying different regions of gene). So which expression vector to choose for representing expression of a gene? I also do not have bead level data for summarizing gene expression for my chosen experiment, as I am using publicly available data.

Please help. Thanks

Answer 1

[Kevin Blighe]

If you are confident that the data has already been normalised (check the distribution via box-and-whiskers and histograms), then, for all intents and purposes, you can use the data for whatever purpose downstream. I see no problem using microarray data for correlation analysis.

The level of probe summarisation will depend on the exact 'chip used. If you must, summarise the data to unique genes via mean or median.

Kevin