Scientists have utilized artificial intelligence to create a strong predictive model that can identify the most effective immune cells for killing cancer. This model can be used in cancer immunotherapies and has been described in the journal Nature Biotechnology. When combined with other algorithms, this model can be used for personalized cancer treatments.
According to Ludwig Lausanne’s Alexandre Harari, the use of artificial intelligence in cellular therapy is a recent development that could revolutionize treatment options for patients. Along with graduate student Rémy Pétremand, Harari led a study that highlights the potential of this approach.
Cellular immunotherapy involves the extraction of immune cells from a patient’s tumor. These cells can then be modified to enhance their natural ability to fight cancer before being reintroduced into the patient’s body after being grown in a laboratory setting. T cells, a type of white blood cell, play a crucial role in this process.blood and patrol for cells that are infected with viruses or cancer.
T cells that enter solid tumors are referred to as tumor-infiltrating lymphocytes, or TILs. However, not all TILs are effective at identifying and attacking tumor cells. Harari explained, “Only a small number are actually able to react to the tumor — the majority are just bystanders. “Our goal was to identify the TILs that have T cell receptors capable of recognizing antigens on the tumor.”
To accomplish this, Harari and his team created a new AI-driven predictive model called TRTpred, which can rank T cell receptors (TCRs) based on their ability to react to the tumor.When developing TRTpred, the researchers utilized 235 TCRs collected from patients with metastatic melanoma, which had already been categorized as either tumor-reactive or non-reactive. The team inputted the global gene-expression profiles of the T cells containing each TCR into a machine learning model in order to detect distinctive patterns that differentiate tumor-reactive T cells from non-reactive ones.
“TRTpred has the ability to learn from one T cell population and establish a rule that can then be applied to a new population,” Harari explained. “So, when presented with a new TCR, the model can analyze its transcriptomic profile and predict whether it is tumor-reactive or not.”
TheThe TRTpred model examined TILs from 42 patients with various types of cancer and successfully identified tumor-reactive TCRs with approximately 90% accuracy. Additionally, the researchers improved their TIL selection process by implementing a secondary algorithmic filter to specifically target tumor-reactive T-cells with “high avidity”, meaning they strongly bind to tumor antigens.
“TRTpred solely predicts whether a TCR is tumor reactive,” Harari stated. “However, certain tumor-reactive TCRs bind particularly strongly to tumor cells, making them highly effective, while others do not.”Only operating in a sluggish manner will not yield desirable results. Identifying the potent binders as opposed to the weak ones is crucial for effectiveness.”
The study illustrated that T cells identified by TRTpred and the secondary algorithm as both tumor-reactive and possessing high avidity were more frequently located within tumors rather than in the surrounding supportive tissue, known as stroma. This discovery is consistent with previous research indicating that efficient T cells typically infiltrate deeply into tumor islets.
The team then implemented a third filter to enhance the recognition of various tumor antigens. “Our goal is to maximize the likelihood that TILs will target a wide range of different antigens.””This final filter categorizes TCRs based on similar physical and chemical characteristics,” explained Harari. The researchers believed that TCRs within each group likely recognize the same antigen. “Therefore, we select one TCR to amplify within each cluster, to increase the likelihood of targeting distinct antigens,” explained Vincent Zoete, a computational scientist at Ludwig Lausanne who created the TCR avidity and TCR clustering algorithms.
The researchers refer to the combination of TRTpred and the algorithmic filters as MixTRTpred. To test their method, Harari’s team grew human tumors in mice.The researchers obtained TCRs from the TILs and utilized the MixTRTpred system to pinpoint T cells that had high avidity, were reactive to tumors, and targeted multiple tumor antigens. They then modified the T cells from the mice to express these TCRs and demonstrated that these cells were capable of eradicating tumors when transferred into the mice.
“This approach holds the potential to address some of the limitations of current TIL-based therapy, particularly for patients with tumors that are unresponsive to existing treatments,” stated George Coukos, the co-author of the study and Director of Ludwig Lausanne, who intends to initiate a Phase I clinical trial.The technology will be tested in patients.
“Our combined efforts will result in an entirely new form of T cell therapy.”
This research was funded by Ludwig Cancer Research, the Swiss National Science Foundation, the Cancera Foundation, the Mats Paulssons Foundation, and the Biltema Foundation.
Alexandre Harari is a principal investigator in the Hi-TIDe team at Ludwig Lausanne and an associate professor at the University of Lausanne.
