Hello, I'm currently using Limma and Limma-Voom to analyze some Microarray and RNA-Seq datasets, with the hope of finding some DEGs in my samples. Farther downstream, I'm looking at some motif analysis regarding some specific transcripts that have particular motifs present. The challenge is that not all transcript variants of a gene necessarily have the motif, so I'm not sure if I can accurately relate my DEGs to genes with the motif. *What are the challenges or problems with feeding Limma (or even EdgeR and other packages) transcript counts instead of gene counts?* I'd feed it a custom library size so it won't double count using rowSums, but would it cause other problems when I get my final list of differentially expressed *transcripts*? I also had some additional technical questions regarding limma: 1. Is the standard practice to input all gene expression counts from an experiment into the same voom/limma objects? That is, I'm looking at different cell types, treatments, etc, and so is it better to keep them separate or to keep them together? 2. My current method also only uses reads that map uniquely to a given gene; that is, it ignores reads that map to ambiguous areas of overlapping genes. I assume that this is also standard practice in calling DEGs? Thanks, Yogesh Yogesh Saletore yos2006@med.cornell.edu http://www.ysaletore.com/ PhD Student Christopher Mason Lab, http://www.masonlab.net Department of Physiology and Biophysics and the Institute for Computational Biomedicine Weill Cornell Medical College of Cornell University 1305 York Ave., New York, NY 10021 [[alternative HTML version deleted]]