I have some gene measurements from a microarray experiment. On each array, I have a set of "non expressed" background genes. I have three replicate measurements for some tissues. How to use the background genes (not normally distributed) to evaluate if the other genes are expressed in each tissue?

  • $\begingroup$ You don’t reinvent this kind of techniques by yourself. You should start by a serious bibliographic research. I don’t know this too much since I never worked with these data, however I once started to look at the biblio. Parameter estimation for the calibration and variance stabilization of microarray data and Variance stabilization applied to microarray data calibration and to the quantification of differential expression, both by Huber et al, seem good starting points. Don’t forget that all proposed methods have been implemented in software packages, don’t reinvent the wheel. $\endgroup$ – Elvis Jan 17 '12 at 14:12
  • $\begingroup$ yeah... Problem is that trad. analysis looks only at diffs. in exp. between groups. I have done this (anova problem) for the tissues, but I would in addition like an estimate of which genes are expressed above background in which tissues. Thought of doing this by comparing each gene to my background distribution on the same array, asking "is this true gene expression value drawn from a distribution with same mean as false background genes?". I could do this, and get a p-value, but the tickle is that I have 3 arrays to ask this question for for each tissue, i.e. I will get 3 pvals for each gene $\endgroup$ – Nightwriter Jan 17 '12 at 14:54
  • $\begingroup$ Hum, yes, this looks fine. In that case you could think to non parametric methods like the bootstrap. Regarding the fact you have replicates, once the data are "normalized" you should be able to pool them (under the assumption they are identical up to the measurement errors). However I am sure you can’t be the first to do that. Did you try scholargoogling "microarray expression profile" ?! $\endgroup$ – Elvis Jan 17 '12 at 15:04

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.