Researchers from Argonne National Laboratory and University of Chicago have developed a new supercomputer-based tool to model tumor progression and destruction by the immune system.
Their work demonstrates that computational simulations of immune-tumor interactions can infer whether or not a given tumor can be destroyed with immunotherapy. This exciting development may one day curate personalized therapies for cancer based on a patient’s individual biology.
Cancer immunotherapies are able to significantly improve survival in various cancer types, however, immunotherapies do not benefit all patients equally: Only 10-20% of patients experience durable, partial, or complete response. Despite continued research, scientists have not been able to reliably determine whether or not a given patient will benefit from immunotherapy. To address this need, the authors of the study applied a computational model to study how cancer and immune cells interact, and how changing various aspects of the model, mimicking the effects of immunotherapy, can lead to cancer regression.The researchers began with a cell simulator, which incorporates realistic growth, death, and heterogeneity of cells. Then, they modelled immune and cancer cell populations. In the model they varied six different variables, including immune cell kill rate, immune cell attachment rate, and immune cell migration. At the end of each simulation they would quantify how many tumor cells remained, and compared this across changes in different variables.
The researchers performed the simulations on the Argonne National Laboratory supercomputer cluster, generating over 500,000 different results by testing various combinations of the parameters. They found that in 80% of cases the immune cells were unable to stop cancer progression, whereas in 2% of cases immune cells could kill 99% of tumor cells. However, they also found that within many different possible scenarios, changing one or more variables could improve cancer cell killing, demonstrating a potential path to therapy based on individual patient biology.
“With this new approach, researchers can use agent-based modeling in more scientifically robust ways,” said Nicholson Collier, computational scientist at Argonne and the University of Chicago that was involved in the study.