Usually when you do a genetic experiment, you have a model of how things work in your head, then you do the experiment and test that model on living organisms.

It's entirely possible that your model is way off and you get results that don't mean anything. With computational models, we can test hypotheses (the models in our heads) much faster (don't have to wait for the worms to grow up) and much cheaper (we can do it on the computers all of us already own).