I am trying to run a differential expression analysis for proteomics data using limma. The data has been quantified and normalized using the LFQ method - the sample distributions are log normal, but with a high median 10e7, ranging from 0 to 1e10
The problem arises with genes that are present in one condition and absent another. Before putting the values into limma I need to log them, which converts 0 - Inf, and drops the genes from the analysis. I tried replacing Inf with 0 or adding pseudocounts, but both things break the model.
Is there any way around this?