paired end gives better quality but it also costs more. So it will depend how you want to balance cost/budget versus quality.
Please don't quote me (!) but based on your aim of differential gene expression to find candidates, personally I feel single end is a good place to start, especially if there will be validation by other methods.
I know this is a bit late, but if you are simply looking for differential gene expression, you are likely to be ok using single end sequencing. As with everything, you get what you pay for- paired end sequencing will always be the "better" option, but for differential expression analysis it is likely not worth the additional cost. For transcriptome profiling however (i.e. isoform and splicing differences) it seems that paired end would be better.