Hello! I am using the package TCGAbiolinks to download and analyze the LIHC dataset from TCGA. I am following the pipeline shown here and here. .
Briefly explained for those who do not wish to read the resources: once the data is downloaded, firstly the function TCGAanalyze_Preprocessing is applied in order to eliminate any potential outliers. Afterward, normalization is performed using the TCGAanalyze_Normalization function, which uses an EDAseq method to normalize by gene length. Then, the TCGAanalyze_Filtering function is applied (the method is 'quantile' and the cut value is set at 0.25, not too sure what this part is about, to be honest). Finally, TCGAanalyze_DEA offers two methods for differential expression analysis: limma and edgeR. The default method is edgeR and it's the one used in both of the links I cited.
My question is: in order to use limma should I follow the exact same workflow and just change the method parameter of the function? or do I have to change anything in the process (normalization, filtering, or anything else) in order to use limma? Sorry if the question is too dumb but I'm a beginner and while I am aware that the limma and edgeR methods are different, I am not sure if the input data should differ or not. Thanks in advance!
I am usually sceptical when package do DE under the hood. I would check whether you can extract a matrix of raw counts from this object you downloaded and then simply analyze manually with edgeR. The manual is very comprehensive, alternatively there are some code suggestions here: Basic normalization, batch correction and visualization of RNA-seq data
The thing is that you usually want to make some exploratory QC, e.g. via PCA to see whether there are obvious outliers that might need to be removed.
I will check that resource, thanks! and yes there is indeed a step of elimination of outliers which I had forgotten to include, my bad! I edited the post to add that step.
It is just a suggestion, you can probably use the method they provide as in-built, I personally prefer to have full control over things though.