Familial hypercholesterolemia (FH) is the most common inherited metabolic disorder in the world with about 2.5 million Europeans affected. In the Netherlands it is estimated that 60,000 people have this condition. Individuals with FH present a high cardiovascular risk but there are different effective treatments to reduce this risk; however, these high-risk individuals are not being clinically identified. An accurate diagnosis through genetic testing confirms the disorder in only about half of the cases, in the other 50% the cause for the severe hypercholesterolemia remains unknown, preventing the best management of their disorder. FH-EARLY will enable new strategies for earlier diagnosis and co-management of familial hypercholesterolemia. This will be facilitated by well-defined data flows, identifying new mechanisms through analyzing big sets of biomolecular data (multiomics), explainable AI modelling and co-creation with families with FH and caregivers.
Preventing coronary heart disease for everyone
FH-EARLY will help prevent premature coronary heart disease, not only for FH patients, but also within the general population: with potential to generate considerable economic and social benefits. FH-EARLY will develop three interrelated solutions for FH: an array for earlier diagnosis, an assay for risk stratification and co-management strategies. This will provide faster and more affordable diagnosis and treatment of FH delivering the right intervention/service to FH patients and families at the right time, starting much earlier in life. FH-EARLY will also help us find new ways to identify novel pathways and mechanisms involved in severe hypercholesterolemia.
The role of CWI and LUMC
Researchers Peter Bosman (head of the Evolutionary Intelligence group at CWI) and Tanja Alderliesten (head of the AI-based Innovations research group at the department of Radiation Oncology of LUMC) will be using explainable AI techniques to predict the risk of an FH patient getting heart disease. “We will receive multiple clinical data sets from our research partners to train our AI models”, Alderliesten explains. “These models have to answer questions like: how will the disorder develop in this individual patient? Who is at risk for a heart attack and who isn’t? And of course: why?”, Bosman adds.
Both researchers were asked to join their European partners because of the explainable AI techniques they have been developing since a few years in their joint ICAI lab ‘Explainable AI for Health’. Bosman: “Our models have the potential to show you immediately what has been learned from the data. This can lead to novel insights, but also to new questions. You might be able to tell from inspecting the model that you need more or different data. It is an interactive and iterative process.”
Header picture: shutterstock.com
