Assistance from AI in finding the cause of illnesses

Headaches, high blood pressure or digestive problems can be signs of many different medical conditions. Often, treatment focuses on the affected organs or organ systems and is aimed at alleviating symptoms.

This is something experts at Friedrich-Alexander-Universität Erlangen-Nürnberg (FAU), including computer scientist Prof. Dr. David B. Blumenthal, are hoping to change. Modern medical research is unlocking an increasing number of mechanisms behind certain diseases on the molecular level, thus facilitating new and more effective approaches in treating these diseases.

"We call this approach network medicine," explains David B. Blumenthal. "It aims to classify diseases in terms of underlying molecular mechanisms instead of defining diseases mainly in terms of symptoms or affected organs."

For example, asthma is related to diabetes mellitus on the molecular level, even though different organs are affected at first glance. By understanding these mechanisms and networks, physicians could develop more targeted and more effective treatments that could, in particular, cure people with complex medical conditions, instead of just alleviating their symptoms.

The key to the new approach could involve medical databases in which medical conditions are linked to proteins, genetic variants, medication, symptoms, and so on. This is where Professor Blumenthal and his team come in. The FAU scientists are experts in biomedical networks and thus also preparing and precisely evaluating medical data using artificial intelligence.

In collaboration with researchers from the University of Hamburg, TU Braunschweig and other international cooperation partners, the FAU team only recently examined important international medical databases to find out how useful they could be for deciphering the molecular mechanisms behind disease.

"Our study has shown that the uncritical use of large publicly-available databases is problematic in the search for disease mechanisms because the organ and symptom-based definitions of diseases distort the search results," says Professor Blumenthal. "To fulfill the promises of network and high-precision medicine, focused studies with molecular data for well characterized patient cohorts remain the gold standard."

He continues, "In close collaboration with immunologists at TU München, we are currently working on AI models with which we hope to identify molecular factors that facilitate or impede the success of immunotherapy in cancer."

Provided by Friedrich–Alexander University Erlangen–Nurnberg