RNA-seq variant calling for somatic

I want to find best practice to do variant calling from this page : https://software.broadinstitute.org/gatk/best-practices/
I have exome-seq and RNA-seq data and I think because the sample is from human cancer, the workflow I need to follow is somatic. I don't see any best practice that use RNA-seq for somatic cell. Any suggestion which best practice I should follow and which data, RNA-seq or exome-seq I should use? Thank you.