Machine-Learning to Bridge Heterogeneity in Clinical Data Resources and Optimize Model Generation
Principal Investigator:
Dr Alexandra Dumitriu
Approved Research ID:
45075
Approval date:
March 26th 2019
Lay summary
In order to impact prevention, diagnosis, and treatment of a wide range of serious illnesses, Sanofi plans to undertake research to characterize longitudinal disease risk profiles. This will be accomplished through machine learning approaches applied to (1) different types of health information collected for the UK Biobank participants and information available from (2) Sanofi's clinical trials and (3) additional Real World Data resources available to Sanofi (e.g. electronic health records and claims). We will generate mathematical representations of disease progression by bridging these three types of data resources. Specifically, models of disease progression that include pre-symptomatic through late-stage disease (data permitting) will be created. We will focus on modeling of conditions such as Asthma, Chronic Obstructive Pulmonary Disease, Cancer, Diabetes in the absence of other metabolic disease, genetically at-risk Parkinson's Disease patients, and other chronic or acute conditions. Outputs will include novel definitions of disease, comprehensive patient profile summaries, and reusable disease models. These models will provide actionable information that can inform patient characterization for physicians and design of more efficient clinical trials, or provide refined biomarkers or disease endpoints to accelerate the drug discovery process.