We are a computational laboratory focused on integrating biological, socio-demographic, and clinical data types to learn about the pathophysiology of psychiatric and neurological illnesses. By developing statistical models of illness risk and prognostic trajectories at the population level, we seek to develop clinically useful tools which may inform optimal treatment strategies or lead to the development of novel therapies.
By combining data from typically siloed fields, we aim to understand mental health as the product of more than just brain structure and function, examining the complex interaction of genetics, environment, and lived experience. In order to develop this whole person perspective, we apply biostatistical and machine learning methods to probe the high dimensional space of "big data" produced by local and international population-based studies (such as the UK Biobank and Canadian Longitudinal Study on Aging) and distill those features which are most informative in making clinical predictions.
We strive to balance complexity and utility, by emphasizing that research in this often abstract space must be made accessible to clinicians and policy makers. We continually remind ourselves that our goals are not merely academic; ultimately, the purpose of our work is to meaningfully improve the lives of those affected by mental illness.
We are located in the Krembil Centre for Neuroinformatics on the 12th floor of the Centre for Addiction and Mental Health (CAMH) in Toronto (250 College Street).
Dan Felsky PhD, Lab Head.
Independent Scientist, Krembil Centre for Neuroinformatics, Centre for Addiction and Mental Health
Assistant Professor of Psychiatry,
Associate Member, Institute of Medical Science,
University of Toronto.
Use hypothesis-free approaches to clustering features and individuals to reveal distinct illness subtypes
Apply high dimensional statistical models using cross-disciplinary data types inclusive of central and peripheral features
Dissect heterogeneity in populations by identifying key biomarkers of illness risk and progression
Use population models to make group- and individual-level predictions of symptom trajectory and treatment response
Use machine learning to prioritize predictive features and better understand the pathophysiology of disease