The biggest difference between academic tools and those in industry is that they're sinking major funds into producing (expensive, hard to produce at scale) training data. That, in theory, should allow them to develop better algorithms for actually predicting which mutations in the tumor are going to be the best (immunogenic) targets. This tool, and several others like it, are modular enough to allow you to plug in whatever prediction algorithm you like, while still getting the benefit of all the other steps.