Comparing Expression Values From Two Tissues (Under 2 Conditions In 9 Subjects)

Question

[Colm]

I have a dataset (human microarray) with a total of 36 samples; 9 subjects x 2 conditions x 2 tissue types.

I want to examine the relationship in expression between the two tissue types under both conditions. I've used co-inertia analysis to get an idea of the overall similarity, along with a very basic correlation of average gene expression across all arrays between the two tissues. I've also started using CCA to look at correlation between variables.

What I'd like to do is find a subset of genes that is highly correlated (or more specifically, shows greatest covariance) between the two tissue types and also a subset that has least covariance.

What would be the best approach for this?

Answer 1

[Sean Davis]

From your data description, it sounds like you have paired samples. Why not simply calculate a correlation between the 18 values from tissue 1 with the paired 18 values for tissue 2? You end with a vector of correlations, one for each feature on the array. You can pick the top X% of those features. If you are interested in a statistic for each feature, you could use permutation or, in R, cor.test to establish some level of statistical significance.