Kinase inhibitors are a successful category of therapeutics used in treating diseases such as cancer, autoimmunity, and neurodegeneration. Despite their efficacy, these drugs often present clinically relevant adverse events that can limit their therapeutic utility or lead to their failure in clinical trials. The adverse event profile of a kinase inhibitor can be explained by its off- and on-target binding profile. Unfortunately, there are limited resources to couple inhibition of a specific kinase to clinical adverse events. Discerning which adverse events can be attributed to a specific kinase, and which are more generally caused by the kinase inhibitor drug class, is crucial for designing next-generation drugs that avoid toxicity and enhance clinical viability. By integrating adverse event incident data from the FDA Adverse Events Reporting Database with machine learning-predicted molecular binding profiles, we developed a statistical method that associates specific adverse events with potent inhibition of certain kinases. We also identify general adverse events inherent to the kinase inhibitor class. We validate our model through an extensive literature review of known kinase-adverse event pairs, comparison with the OnSIDES drug label side effect dataset, and prospective prediction of adverse events of recently approved kinase inhibitors. We show that our method can recapitulate well-established kinase-toxicity associations and identify previously unreported kinases associated with adverse events.