Approved research

Identification of multi-modal biomarker signatures in psychiatry using artificial intelligence

Lay summary

Approximately 1 in every 8 individuals worldwide live with a mental disorder, which encompasses substantial disruptions in thinking, emotional regulation, or behavior. These mental disorders manifest in various forms, and although effective prevention and treatment options are available, a significant portion of the population lacks access to these commendable care services. Schizophrenia is a severe mental illness affecting approximately 1% of the population. With its early onset and typically chronic disease course, and the limited effectiveness of current therapeutic approaches, it is one of the most debilitating illnesses. At the same time, patients with schizophrenia show a strongly increased risk for type 2 diabetes and cardiovascular illness, which contributes to the massively increased excess mortality compared to the general population. The current diagnostic process still relies on a largely subjective evaluation of symptom constellations that strongly overlap with those of other disorders. Similarly, the choice of the most appropriate therapies can currently not be guided by an in-depth understanding of the biological disease process in an individual patient. In this project, we aim to use artificial intelligence approaches for the identification of biomarker signatures that provide deeper insights into the biological basis for schizophrenia and related illnesses. We will explore whether clinical, neuroimaging, genetic and other features can be analyzed to identify algorithms that inform us at the individual participant level about affected biological processes. For this, we will particularly build on modern forms of artificial intelligence approaches, such as the so-called "large language models". Furthermore, we will use our existing approaches to characterize genetic risk effects and explore associations with data types relevant for schizophrenia and related conditions in the UKB cohort. Finally, we will explore the development of biomarker patterns across the lifespan, which will help in the identification of disease trajectories, as well as with pinpointing risk-relevant age periods. Such in-depth understanding of biological processes and their trajectories will provide a promising basis for the future development of personalized medicine approaches, where diagnosis and therapy selection can be informed by measures relevant to the individual person.