Hi,
There are different methods in handling microarray data, such as SAM, Rank Products, limma, or simple student test... I am just wondering which methods do you prefer in finding differently changed genes? Personally, I use student test or anova, sometimes with SAM.
Thanks,
Limma is the Α and the Ω in my book.
We sometimes use SAM. On our last expt., involving miRNAs, the differentially expressed genes were detected using a simple T test. We had just three replicates.
t-test with three replicates is not applicable. Some distributions have to be calculated and with 3 points, it's not possible. You 'd rather use a wilcoxon (non-parametric) test whn dealing with small datasets.
Well, I cannot argue on my colleague's behalf. That said, in brief, with his 3 expression data points per microRNA he estimated the variance, and combined with n=3 ran the T test. He acknowledges that this is not ideal but states that it is indeed possible.
When I deal with data from on 3 samples experiments (shotgun ms data), I used PLGEM. The results seemed fine, so it may be an alternative.
Anova with a FDR
in your case of small number of replicates, i think RankProducts (non-parametric test) would be more suiable. However, there is no goldstandard test if you have enough time why not try a number of parametric tools and non-parametric tools and then take the intersect of both which ultimatel that most reliable source of significant differently expressed genes.
Log in to answer this question.