0
$\begingroup$

Need advice: how to approach discrepancy in differential microarray gene expression test results: what to do if ANOVA, ttest, SAM and Limma procedures show different results and especially more discrepant when using log transformations and normalization operations?

$\endgroup$
1
$\begingroup$

I wouldn't expect different methods to give the same results. Further, why are you even testing non-normalized datasets, the results of that are completely and utterly useless for any purpose other than showing that normalization is important. In addition, an T-test is a special case of an ANOVA (and of course limma is itself using a moderated T-test, though it's going to have significantly more power than the others), so I have to ask myself exactly what sort of design you're using that it's appropriate to substitute one for the other and still get apparently vastly different results.

In general, I would strongly encourage you to work with a local bioinformatician or statistician for what is presumably your first microarray analysis, particularly if you don't have a strong statistics background.

$\endgroup$
  • $\begingroup$ We are all academically educated in medicine and computer science, you know. However, we did 7 years of angiogenesis, not microarray research and started this theme recently, We have read many books, and advices are quite controversial. In routine, we do quantile normalization and log2 or log1o transformation for visualization. Design is simple: 4 columns of control bioichips and 6 of experimental data. Problem is that no up- or down- significant expression is achieved when using massive normalization, log transform, FDR and so on. Thank you for helping me. $\endgroup$ – VassiaAlk Jan 17 '15 at 9:55

Your Answer

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

Not the answer you're looking for? Browse other questions tagged or ask your own question.