Ludwig Cancer Research scientists have developed a predictive model that uses artificial intelligence to identify effective cancer-killing immune cells for use in cancer immunotherapies. This model, called TRTpred, can be applied to personalized cancer treatments tailored to the unique cellular makeup of each patient’s tumors. Cellular immunotherapy involves extracting immune cells from a patient’s tumor, potentially enhancing their abilities, and reintroducing them after expansion in culture. T cells are the main type of lymphocyte that patrol the body looking for infected or cancerous cells, specifically tumor-infiltrating lymphocytes (TILs) that can recognize and attack tumor cells.
To develop the TRTpred model, 235 T cell receptors (TCRs) from patients with metastatic melanoma were analyzed and identified as either tumor-reactive or non-reactive. The team used machine learning to analyze global gene-expression profiles of the T cells with each TCR to determine patterns that differentiate tumor-reactive T cells from inactive ones. The TRTpred model can predict whether a TCR is tumor reactive based on its transcriptomic profile by learning from one T cell population and applying that knowledge to a new population. This model was shown to accurately identify tumor-reactive TCRs with about 90 percent accuracy in patients with melanoma and other cancers.
In addition to predicting tumor reactivity, a secondary algorithm was used to identify TILs with high avidity, or those that bind strongly to tumor antigens. T cells flagged by both TRTpred and the secondary algorithm were found to be more embedded within tumors rather than in the surrounding tissue, indicating their effectiveness. A third filter was added to maximize recognition of diverse tumor antigens by organizing TCRs into clusters based on similar characteristics, with the goal of targeting multiple antigens. This combination of TRTpred and the algorithmic filters is called MixTRTpred.
To validate the approach, TCRs were extracted from human tumors in mice, and T cells were engineered to express TCRs identified by the MixTRTpred system as tumor-reactive, high-avidity, and targeting multiple antigens. The engineered T cells were able to eliminate tumors in the mice, demonstrating the potential of this method to improve TIL-based therapy for patients whose tumors do not respond to current treatments. A Phase I clinical trial is being planned to test this technology in patients, offering new possibilities for T cell therapy in the future.
This study was supported by various research foundations and institutions, and the researchers involved have backgrounds in computational science, immunology, and cancer research. The development of the TRTpred model and its application in identifying effective TILs for cancer immunotherapies represent a promising advancement in personalized cancer treatment. The use of artificial intelligence in cellular therapy offers new clinical options for patients, with the potential to overcome limitations of current TIL-based therapies. The combination of predictive models and algorithmic filters could revolutionize T cell therapy by maximizing the effectiveness of immune cells in targeting various tumor antigens.
