ExTrAct-AML

Precision medicine aims at delivering the best treatment to an individual patient based on their unique characteristics. In recent years precision medicine approaches have largely been reliant on characterizing patients genomes to identify promising medicines for individuals and defined patient subgroups. These approaches have led to selected success stories. However, childhood cancers have not yet profited as much from such approaches, which is mainly due to their rarity and their distinct genetic profiles. One particular disease in need of more personalized treatments is pediatric acute myeloid leukemia (pedAML) a rare childhood blood cancer with poor outcome. Survival of pedAML patients has increased drastically over the recent decades by optimizing treatment in large international trials, but the improvements in survival rates have slowed down in recent years. Treating pedAML patients based on their genetic profiles can only partially improve this situation, because there are no targeting medicines for most mutations in this disease. To tackle this challenge and ultimately improve outcome in pedAML, we set up a scientific partnership between the St. Anna Childrens Cancer Research Institute and the CeMM Research Center for Molecular Medicine of the Austrian Academy of Sciences. In this team, we aim to perform comprehensive ex-vivo functional profiling of pedAML patients and their tumors. It allows us to test a many different medicines on individual patients by performing these tests on the tumor outside of the patients body. By additionally profiling the genetic makeup of the tumor and its internal signaling pathways, we aim to create a comprehensive map of promising medicines in pedAML where the genetic and signaling information allows us to understand why these medicines may work in individual patients and specific subgroups. We aim to first perform this profiling on samples from patients diagnosis, but then also characterize how these profiles change over time as patients are treated. With this approach, we hope to better understand the causes of poor treatment response in individual patients and specific subgroups. Additionally, we aim to establish this characterization as a tool to identify patients at risk early on and to be able to assign high-risk patients to promising trials early on, to improve their chances of survival.

Pediatric acute myeloid leukemia (AML) remains one of the most aggressive childhood cancers, and despite therapeutic progress, many children either respond poorly or relapse. To help address this, we have developed a new image based screening platform that allows us to study how each patient's leukemia cells react to a wide range of clinically relevant drugs-before treatment even begins. Using high resolution imaging and artificial intelligence, we analyzed hundreds of thousands of cells from each patient sample. This enables us to create detailed drug response profiles that reveal which medicines the leukemia cells are sensitive or resistant to. Unlike genetic testing alone, our approach captures how the cancer behaves functionally, providing information that is directly relevant for treatment decisions. We found that these drug response patterns can predict important clinical outcomes right at diagnosis. Even with simple machine learning models, we were able to identify patients who belong to higher risk groups and to forecast whether leukemia would still be detectable after the first cycle of therapy. This suggests that functional testing could support more precise and timely risk assessment. Our results also provide insight into why treatment responses differ between patients. Leukemia cells that appear more differentiated (monocytic like) responded better to certain chemotherapy agents, while stem cell like leukemia cells-typically more resistant-showed vulnerabilities to targeted approaches such as BCL2 or HDAC inhibitors. These findings highlight cell state specific weaknesses that may guide individualized treatment strategies. Building on previous work, we also identified a potentially new subtype of pediatric AML characterized by co occurring BCR::ABL1 and CBFA2T3::GLIS2 fusions-an important step toward refining diagnosis and risk categorization. Several follow up efforts are already in progress. We are continuing to profile high risk pediatric AML samples to deepen our understanding of therapy resistance. At the same time, we are coordinating efforts to integrate this functional screening method into the high risk arm of the ongoing AML BFM 2020 trial. This will allow us to test the method prospectively at diagnosis and compare its predictions directly with patient outcomes in real time. While larger studies are needed to fully validate our findings, our work demonstrates the strong potential of functional precision medicine: by identifying high risk patients early and revealing personalized therapeutic vulnerabilities, we aim to improve treatment outcomes and reduce relapse rates for children with AML.