Hello. I have a technical question.. From data available in some online forums, it seams that edgeR can be used to analyze continuous data sets. My experimental design does not include any "treatments" nor "replicates". Its simply a gene expression counts along an environmental gradient (with more than 200 samples, spanning along salinity gradient, including some zero salinity measurements). Should I bin the data into discrete values with "replicates" of similar doses, or edgeR can analyze this data set without any replicates? I am searching for salinity responsive genes, and I wish to distinguish expression pattern along different salinity doses. If so, should I include the salinity level as "grouping" factor in edgeR analysis?

You could bin salinity into groups if you want, but there is no need to do so. I would personally use edgeR to fit a smooth expression curve for each gene as a function of salinity level. See Section 9.6.2 of the limma User's Guide for a suggestion of how to use regression splines for this purpose. The same approach works in edgeR as in limma.

Seel also https://support.bioconductor.org/p/72282/