On an RNA-Seq dataset, I want to apply a clustering algorithm which requires zero-mean uni-variance gene expression levels. I am wondering whether it makes sense to use "scale" function in R after taking logarithm of the counts. For microarray data, I was actually using the scale function which (as far as I know) makes sense; but this is the first time I am using RNA-Seq. I heard about "voom" normalization (in "limma" library on CRAN), but I don't really want to use that since I am not really familiar with the details.

  • $\begingroup$ What are the datasets; different samples? The analyses and the statistical procedures that you use depend on what kind of data you have. There are several (infinite) distributions that can have 0 mean and 1 variance. The question is, what is your expected distribution i.e. the null model. Certain data are known to follow certain models. So the question boils down to what is your data. $\endgroup$ – WYSIWYG Aug 2 '16 at 8:13
  • $\begingroup$ My data is gene expression data and the algorithm assumes a zero-mean uni-variance Gaussian distribution. $\endgroup$ – user5054 Aug 2 '16 at 13:30
  • $\begingroup$ So you just have one sample? Technically, the different genes are different random variables. Moments like mean and variance are for one random variable. Comparing different genes is like comparing apples and oranges. Moreover, the expression of all genes is not expected to follow gaussian. For more details see this post. $\endgroup$ – WYSIWYG Aug 3 '16 at 13:22

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.