Cancers are among the most deadly and intractable of human diseases, for which the standard of care may involve aggressive yet only partially effective therapies. Immunotherapy, which harnesses the body's immune system to target cancer cells and can lead to long-term remission in some of the most advanced treatment-refractory cases, and liquid biopsy assays, which can potentially detect cancer at an earlier and more easily treatable stage, require the identification of cancer-specific genetic variants. Overall in this work, I show the potential scale of sample counts and types required for a comprehensive normal background of splicing against which truly cancer-specific RNA splicing can be identified; the instability of such identifications and their sensitivity to specific methods and filtering parameter values used; and the poor reliability of intron retention detection from short-read RNA-seq data.