Artificial intelligence to help tumor immunology

Cancer treatment has been revolutionized in the past decade by new methods of immunotherapy. This involves not attacking a tumor directly, but rather using the existing cells of the immune system. These are usually able to recognize and eliminate malignant tumor cells. However, many tumors have the ability to prevent or severely limit an effective immune response. Immunotherapy aims to restore the misguided immune system's ability to recognize and destroy tumor cells.

Immunotherapy against cancer is not successful in all patients. Resistance to cancer immunotherapies has been shown to be frequently associated with tumor microenvironment (TME) composition. In oncology, the properties of the TME are already being used as biomarkers to make predictions about how a cancer will develop. This is done using imaging techniques that map the type and location of individual cells within the TME. Patterns of gigantic cell assemblies emerge, which in their totality and structure influence the success or failure of cancer immunotherapy. How exactly this works, however, remains elusive.

"New high-resolution imaging techniques have shown that disease mechanisms are indeed related to details of the spatial arrangement of specific cell types in tissues," notes Prof. Kevin Thurley of the Institute of Experimental Oncology at the University Hospital Bonn. "Using a combination of mathematical modeling and artificial intelligence methods, we will investigate these phenomena in detail, in direct collaboration with experimental and clinical research at our University Hospital."

Artificial intelligence (AI)-based methods for image analysis are already well advanced today. The situation is different when simulating complex systems, due to the large number of interacting cells within a tissue. Given the multitude of cell types involved, the different cellular processes taking place there, and the complex tissue architecture, such a simulation requires extremely high computational resources. However, it can help to simulate the TME of a tumor and thus draw conclusions about tumor development.

The overall goal of "InterpretTME" is to develop interpretable machine learning (ML) methods for studying complex cellular systems. These are to be used to gain insights into the nature of TMEs. "Machine learning is already used in many places in the hospital to process image data," explains Prof. Jan Hasenauer, of the Life & Medical Sciences Institute (LIMES) at the University of Bonn. "We will go one step further and investigate the extent to which information about mechanisms can also be obtained." One aim is to investigate the role that individual immune cell types present in the TME play in the development of different tumor types. In addition, the researchers want to determine what effect chemotherapeutic agents and biological drugs have on the TME of different tumor types. Prof. Michael Hölzel and Prof. Marieta Toma from the University Hospital Bonn and Prof. Alexander Effland from the University of Bonn are also involved in the project.

MEDICA-tradefair.com; Source: Rheinische Friedrich-Wilhelms-Universität Bonn