Intensity Of A Gene Signature In A Microarray?

Question

[User 7365]

How could I compare the "intensity" of a gene signature between multiple microarrays? For instance, upon heat stress in yeast, ~300 yeast are up-regulated and ~600 genes are down-regulated. If I wanted to claim that the intensity of the heat shock response is greater at 42 than at 37, greater at 37 than at 35, greater at 35 than 33, etc... how could I do that?

Answer 1

[Neilfws]

You seem to have the procedure backwards. We don't go from knowing that genes are up/down-regulated to making claims about their intensities - we do it the other way around.

Very briefly, you would:

1. Generate or obtain the appropriate microarray data. Raw data usually comes as some binary file format, such as CEL files for Affymetrix arrays.
2. Process the raw data. Typically this involves extracting the raw intensities from the binary file, normalising them across arrays and making some statistical corrections for factors such as non-specific background binding of probes.
3. Perform appropriate statistical tests to look for significant differences between conditions (42 versus 37, 37 versus 35 and so on). There are many such tests available, most of which return a list of differentially-expressed genes with some sort of confidence value, such as a p-value.

The Bioconductor software suite is widely used for this type of work. Here are some sample workflows for oligonucleotide arrays.

Answer 2

[User 7365]

You've misunderstood. We know that ~900 genes constitute a stress response in yeast (Gasch et al., 2000). I've tried some new stresses, and some new mutations, which also appear to cause a stress response. I want to be able to make claims about which stress responses are more "intense", ie, treatment with chemical X causes a more pronounced stress response than treatment with chemical Y which causes a more pronounced stress response than is observed in mutant Z, etc etc etc.

[Qdjm]

It's better to add your comment to Neil's answer rather than start a new one because otherwise there's no connection between the two. That said, what's wrong with his answer? If the same genes are differentially expressed in 42 vs 37 and 37 vs 35, and they are differentially expressed in the same direction, that would seem to me to be an indicator of a more "intense" stress response.

[Neilfws]

Yes, this should be a comment, not a new answer. But now a discussion has started here, so I will leave it this time. The problem is your use of the phrase "intensity" of a gene signature". To most people, this implies the raw signal from the microarray. I understand your question but really, my answer stands. The correct approach to this problem requires statistical testing based on normalised array data.