I'm working with RNAseq data from one patient (2 conditions), you recommend analyzing all samples together, because it will boost the discovery power, why ?
That's not what we recommend for RNAseq... that's for germline DNAseq and it's explained in our workshop presentations, as well as in this overview video.
Ok, thank you, I had looked at this post (http://gatkforums.broadinstitute.org/gatk/discussion/5015/merge-multi-sample-bam-files), and it was not specified if it was DNAseq or RNAseq...
So for the RNAseq, you advise me to use GATK on each of my conditions, individually, and then I compare the variants?
Yes. Have a look at the Best Practices and methods article.