In the journal Nature Biotechnology, scientists at Ludwig Cancer Research reported the development of an AI model to identify the most effective immune cells for cancer immunotherapy. This approach could personalize treatments by selecting the strongest cancer-killing cells from each patient's tumor.
Image Credit: Komsan Loonprom/Shutterstock.com
In conjunction with supplementary algorithms, the predictive model holds promise for personalized cancer treatments. These therapies are designed to customize treatment strategies based on the distinctive cellular composition of individual patients' tumors.
The implementation of artificial intelligence in cellular therapy is new and maybe a game-changer, offering new clinical options to patients.
Alexandre Harari, Department of Oncology, Ludwig Institute for Cancer Research
Cellular immunotherapy is a multi-step process. Initially, immune cells are harvested from a patient's tumor. Optionally, these cells are genetically modified to bolster their inherent anti-cancer capabilities before being cultured to proliferate. Following this, they are reintroduced into the patient's body. T cells, a subset of white blood cells or lymphocytes, play a pivotal role in this therapy, as they scour the bloodstream for cancerous or virally infected cells.
Among T cells, those that infiltrate solid tumors are termed tumor-infiltrating lymphocytes (TILs). However, not all TILs exhibit proficiency in identifying and eradicating tumor cells.
Only a fraction is in fact tumor reactive—the majority are bystanders, the challenge we set for ourselves was to identify the few TILs that are equipped with T cell receptors able to recognize antigens on the tumor.
Alexandre Harari, Department of Oncology, Ludwig Institute for Cancer Research
To achieve this, Harari and his team developed a pioneering AI-driven predictive model known as TRTpred, specifically designed to evaluate T cell receptors (TCRs) and assign them rankings based on their responsiveness to tumors.
Drawing from a dataset comprising 235 TCRs collected from patients diagnosed with metastatic melanoma, previously categorized as either tumor-reactive or non-reactive, the team embarked on training a machine-learning model. This model was trained using the transcriptomic profiles of T cells carrying each TCR, with the aim of discerning distinctive patterns that differentiate tumor-reactive T cells from their inactive counterparts.
TRTpred can learn from one T cell population and create a rule which can then be applied to a new population, so, when faced with a new TCR, the model can read its transcriptomic profile and predict whether it is tumor-reactive or not.
Alexandre Harari, Department of Oncology, Ludwig Institute for Cancer Research
The TRTpred model analyzed TILs from 42 patients with melanoma and gastrointestinal, lung, and breast cancer, identifying tumor-reactive TCRs with approximately 90 percent accuracy. The researchers further refined their TIL selection process by applying a secondary algorithmic filter to screen for only those tumor-reactive T-cells with “high avidity”—that is, those binding strongly to tumor antigens.
Alexandre Harari adds, “TRTpred is exclusively a predictor of whether a TCR is tumor-reactive or not, but some tumor-reactive TCRs bind very strongly to tumor cells and are therefore very effective, while others only do so in a lazy way. Distinguishing the strong binders from the weak ones translates into efficacy.”
The researchers showed that T cells were more frequently identified embedded within tumors than in the nearby supporting tissue, or stroma, after being flagged by TRTpred and the secondary algorithm as both tumor-reactive and having high avidity. This result is consistent with recent studies that demonstrate that tumor islets are often deeply penetrated by effective T lymphocytes.
Subsequently, the group added a third filter to enhance the identification of various tumor antigens.
Harari said, “What we want is to maximize the chances the TILs will target as many different antigens as possible.”
TCRs are grouped together by this last filter according to comparable chemical and physical properties. TCRs in each cluster, the researchers surmised, detect the same antigen.
Vincent Zoete, Computational Scientist at Ludwig Lausanne, said, “So, we pick within each cluster one TCR to amplify so that we maximize the chances of distinct antigen targets.”
Zoete also developed the TCR avidity and the TCR clustering algorithms.
MixTRTpred is the term given by the researchers to the combination of TRTpred and the algorithmic filters.
To validate their approach, Harari’s team cultivated human tumors in mice, extracted TCRs from their TILs, and utilized the MixTRTpred system to identify T cells exhibiting tumor reactivity, high avidity, and targeting multiple tumor antigens. Subsequently, they engineered T cells from the mice to express these identified TCRs and demonstrated their efficacy in tumor eradication upon transfer into the mice.
George Coukos study Co-Author and Director of Ludwig Lausanne, said, “This method promises to overcome some of the shortcomings of current TIL based therapy, especially for patients dealing with tumors not responding to such therapies today, our joint efforts will bring forth a completely new type of T cell therapy.”
Coukos is also planning to launch a Phase I clinical trial that will test the technology in patients.
The research was funded by Ludwig Cancer Research, the Swiss National Science Foundation, the Cancera Foundation, the Mats Paulssons Foundation, and the Biltema Foundation.
Journal Reference:
Pétremand, R., et al. (2024) Identification of clinically relevant T cell receptors for personalized T cell therapy using combinatorial algorithms. Nature Biotechnology. doi.org/10.1038/s41587-024-02232-0
Source: https://www.ludwigcancerresearch.org/