Scientists have developed an AI tool to identify tumor-killing cells with high precision

Using artificial intelligence (AI), scientists from the Ludwig International Cancer Center have developed a powerful predictive model to identify the most powerful tumor-killing immune cells that can be used in cancer immunotherapies, reports News.ro.

Combined with additional algorithms, the prediction model, described in the current issue of Nature Biotechnology, can be applied to personalized cancer treatments that tailor therapy based on the unique cellular makeup of each patient’s tumors.

“The implementation of artificial intelligence in cell therapy is novel and can be a game-changer, offering new clinical options for patients,” Alexandre Harari, who led the study, said in a statement from the Ludwig Center Lausanne.

Cellular immunotherapy involves extracting immune cells from a patient’s tumor, possibly modifying them through genetic engineering to enhance their natural cancer-fighting abilities, and reintroducing them into the body after they have multiplied.

T cells are one of two main types of white blood cells, or lymphocytes, that circulate in the blood and patrol to find virally infected or cancerous cells.

T cells that infiltrate solid tumors are known as tumor-infiltrating lymphocytes, or TILs. However, not all of these TILs are effective in recognizing and attacking tumor cells.

“Only a fraction is, in fact, reactive to tumors,” explained Harari. “The challenge we set ourselves was to identify those few TILs that are equipped with T-cell receptors capable of recognizing antigens on the tumor.”

To do this, the team developed a new predictive model based on artificial intelligence, called TRTpred, which can classify T-cell receptors (TCRs) according to their tumor reactivity. To develop TRTpred, they used 235 TCRs collected from patients with metastatic melanoma, already classified as either tumor-reactive or non-reactive.

The team loaded the gene expression – or transcriptomic – profiles of T cells bearing each TCR into a machine learning model to identify patterns that differentiate tumor-reactive T cells from inactive ones.

“TRTpred can learn from a population of T cells and create a rule that can then be applied to a new population,” explained the researcher. “Thus, when faced with a new TCR, the model can read its transcriptomic profile and estimate whether or not the cell is reactive to tumors.”

The TRTpred model analyzed TILs from 42 melanoma and gastrointestinal, lung, and breast cancer patients and identified tumor-reactive TCRs with approximately 90% accuracy.

The researchers further refined their selection process for these TILs by applying a secondary algorithmic filter to screen for only those tumor-reactive T cells with “high avidity”—that is, those that bind strongly to tumor antigens.

“TRTpred is exclusively a predictor of whether or not a TCR is reactive in tumors,” explained Harari.

But some tumor-reactive TCRs bind very strongly to tumor cells and are therefore very effective, while others do so only in a lazy way. Distinguishing those who bind strongly from those who bind weakly translates into effectiveness, he explained

The researchers demonstrated that T cells flagged by TRTpred and the secondary algorithm as both tumor-reactive and with high avidity were more often found embedded in tumors than in the adjacent supportive tissue known as stroma. This finding aligns with other research showing that effective T cells typically penetrate deep into tumor islets.

The team then introduced a third filter to maximize recognition of various tumor antigens.

“What we want is to maximize the chances that these TILs target as many different antigens as possible,” Harari said.

This last filter organizes TCRs into groups based on similar physical and chemical characteristics. The researchers hypothesized that these TCRs from each group recognize the same antigen.

Thus, the researchers choose within each cluster a TCR to amplify, so as to maximize the chances of having distinct antigenic targets.

They call the combination of TRTpred and algorithmic filters MixTRTpred.

To validate their approach, the team grew human tumors in mice, extracted T-cell receptors (TCRs) from their tumor-infiltrating cells (TILs), and used the MixTRTpred system to identify those T cells that were reactive to the tumors , had a high avidity and targeted several tumor antigens.

They then engineered T cells from mice to express those TCRs and showed that these cells could eliminate tumors when transferred into mice.

“This method promises to overcome some of the shortcomings of current TIL-based therapy, especially for patients facing tumors that are currently unresponsive to such therapies,” said director of the Ludwig Center in Lausanne, George Coukos, co-author of the study.

Now the team plans to launch a phase I clinical trial that will test the technology on patients, which could lead to an entirely new type of T-cell therapy.