Are there any publicly available RNA-seq datasets that have a large number of samples per treatment. I am looking for a training set to validate a machine learning algorithm with so large differences in expression would be best (perhaps multiple cell types?).
One suggestion would be http://www.geuvadis.org/web/geuvadis contains quite a lot of samples (If I'm not mistaken ~400 individuals) (not sure what you mean with large dataset). Also ENCODE has data available, http://genome.crg.es/~jlagarde/encode_RNA_dashboard/. Alternatively you could search for a suitable dataset yourself, e.g. through http://sra.dnanexus.com/