Categorizing and classifying diseases is the bedrock of medical research. Defining a disease in terms of behavior and outcomes leads to researching risk factors, discovering mechanisms, creating therapeutics, and developing treatment plans. This research model has produced a wealth of improvements in modern healthcare, but there are critical exceptions; in particular, heterogeneous diseases – diseases like cancer, Alzheimer’s, diabetes, and sepsis – cannot be classified according to shared mechanisms because each patients’ disease manifests and progresses according to a unique, patient-specific mechanism.

In these cases, we need a model of disease and a treatment plan that are personalized to each patient but going through the entire process of research and discovery for everyone is infeasible. We require new methods for personalized disease modeling, new meta-science disciplines to do research outside of the traditional one-disease-one-treatment setting, and new businesses that bring these discoveries and tools to afflicted individuals.

These needs, combined with the emergence of machine learning paradigms like meta-learning and multi-task learning that formalize the personalization objective, speak to a rich and relatively untapped area of innovation. I see my career as a machine learning researcher, a computational biologist, and a mentor as an opportunity to explore a truly individualized approach to healthcare and train others to grow this budding discipline.