The decreasing cost of genome sequencing technology has lead to an explosion of informa- tion about which mutations are frequently observed in cancer, demonstrating an important role in cancer progression for kinase domain mutations. Many therapies have been devel- oped that target mutations in kinase proteins that lead to constitutive activation. However, a growing body of evidence points to the serious dangers of many kinase ATP competitive inhibitors leading to paradoxical activation in non-constitutively active proteins. The large number of observed mutations and the critical need to only treat patients harboring activat- ing mutations with targeted therapies raises the question of how to classify the thousands of mutations that have been observed. We start with an in depth look at the state of knowl- edge of the distribution and effects of kinase mutations. We then report on computational methods to understand and predict the effects of kinase domain mutations. Using molecular dynamics simulations of mutant kinases, we show that there is a switch-like network of la- bile hydrogen bonds that are often perturbed in activating mutations. This is paired with a description of a software platform that has been developed to streamline the execution and analysis of molecular dynamics simulations. We conclude by examining a machine learning method to demonstrate what kinds information derived from protein sequence alone have the most value in distinguishing activating and non-activating mutations.