Personalized Decision Making
Traditional approaches tell you whether you should do something or not, treat or not treat. But each person has unique risk and reward preferences. Some people are willing to trade off uncomfortable side effects for longer life, others don’t. Our approach takes those differences into account.
There are over 7,000 rare diseases affecting more than 350 million people worldwide and only 5% of these diseases have an FDA-approved drug. In rare diseases, traditional models fail because there are not enough people in the sample to estimate a model. We use an approach that incorporates information not contained in the data sample, such as the mechanism of drug action, data from related trials, and data from adjacent populations.
Immuno-oncology is a new approach to fighting cancer that, instead of targeting the cells directly, attempts to activate the body's immune system against the cancer. Since these therapies are often tested in trials with small populations and we don't always know a priori why some patients respond and others don't, traditional methods of assessing safety and efficacy are no longer adequate. By using patient level submodels, we are able make inferences and predictions with proper uncertainty intervals.
The physiology of the human brain is not completely understood but state-of-the-art research in brain dynamics allows us to fit generative models at the individual patient level and infer parameters associated with the spread of epileptic seizures.