Comparing Gene Expression Profiles Using Z-Scores

Question

[Tonig]

I've got the results of chip-on-chip experiment, a sorted list of genes using z-score, I want to compare it with expression profiles, but I don't know how. I mean, it is possible to get the results from GEO as sorted lists and then compare them? My hypothesis is as follows: In gene expression experiments, the most expressed genes will be those that are in the top of my list. I know ProfileChaser, but it works using Gene expression values not Z-scores. Any ideas?

Thanks

Answer 1

[seidel]

Your question isn't quite clear enough. You want to compare your ChIP chip results with ANY gene expression experiment from GEO? (or some related experiment - just not done in your lab). Also, when you say "In gene expression experiments, the most expressed genes will be those that are in the top of my list...", this is confusing because gene expression experiments are typically of two varieties absolute expression levels and relative expression levels. Thus "the most expressed gene" can be viewed as either the most abundant transcript in a sample, or a gene which changes a lot under a given condition (i.e. gene expression ratios) and may or may not be very abundant. Which are you referring to?

It sounds like you want to know if your sorted list of top ChIP chip genes is enriched for top genes returned by a gene expression experiment, like looking at a venn diagram of the overlap of two lists. If that's the case you can use the hypergeometric distribution to evaluate the significance of any overlap. You have a universe of genes, you have a pool with high z-scores defined by one experiment, and you want to know if those same genes would also rank highly in another experiment. The phyper function in R would be one way to check the overlap.

[Tonig]

Thanks Seidel, my idea is to compare my list with one similar expression experiment in GEO, basically what I want is the list of genes returned by this expression experiment. What I don't understand quite well is the procedure you try to show me: I must get the list of genes coming from expression experiment and then make an hypergeometric distribution analysis to see the overlapping? So, probably I need to download the raw data of GEO and then analyse it?

Answer 2

[Ian]

A similar question was recently asked about determining the probability that two gene lists would overlap by chance:

Probability of gene list overlap

So i guess you could have a list of ChIP-chip related genes (based on a Zscore threshold) and a gene list from the GEO experiment. The only trouble is making sure that the gene nomenclatures are the same in both data sets.