Open science to advance personalized medicine : Leveraging open data via domain adaptation

Personalized medicine promises to overhaul our understanding of health and disease; however, the signals we collect to support it are noisy, variable, and limited. Traditional statistical learning methods are ill-suited to these data-limited regimes, particularly as we advance towards learning in "the single person limit." In this talk, I highlight how drawing on existing, open resources---such as openly shared datasets---can help to set informative priors for inference at the individual patient level. Specifically, I introduce my work using advanced statistical and machine learning methods to unlock opportunities for data re-use, with a focus on human functional magnetic resonance imaging (fMRI) datasets.