Limma differential expression analysis with median-centered data

Question

[jfrigola1]

Hi all,

I've just received a matrix of RNA_seq data (genes at the rows, samples at the columns and values at each cell), where the values at the cells are median-centered values. I would like to know if it's possible to perform a differential expression analysis using limma with this data.

Thanks in advance,

Joan

[Devon Ryan]

It'd be simpler if whomever gave you the dataset could simply provide you with either the raw or estimated counts. Then you don't have to monkey around with things as much.

Answer 1

[Manvendra Singh]

There is a way to do with limma as its written in the limma manual but for that you would need to have replicates and you can group them to perform the analysis.

I would suggest to compute pair wise comparison of median-centered values of genes as their row wise Z-score (expression value subtracted by mean of its row values and divided by its standard deviation). This is easy also on R.

[pedromilanez]

Sorry for the very late comment, after almost 3 years. Do you mean something like this?

exprs.gse <- t(scale(t(exprs(gse))))

After that you can input exprs.gse in lmFit, followed by eBayes?

PS: The data I am working with is from microarray and have been "log2 transformed (signal-to-noise < 3 or flag values >8191 set to missing), centered and scaled, missing values imputed with k-nearest neighbors imputation, batch-adjusted using an ANOVA-based algorithm (pamr) and mean-centered within-array (column wise)."