AI Predicts Tumor-Killing Cells with High Accuracy

Using artificial intelligence, Ludwig Cancer Research scientists have developed a powerful predictive model for identifying the most potent cancer killing immune cells for use in cancer immunotherapies.

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

"The implementation of artificial intelligence in cellular therapy is new and may be a game-changer, offering new clinical options to patients," said Ludwig Lausanne’s Alexandre Harari, who led the study with graduate student Rémy Pétremand.

Cellular immunotherapy involves extracting immune cells from a patient's tumor, optionally engineering them to enhance their natural abilities to combat cancer and reintroducing them to the body after they’ve been expanded in culture. T cells are one of the two main types of white blood cells, or lymphocytes, that circulate in the blood and patrol for virally infected or cancerous cells.

T cells that penetrate solid tumors are known as tumor-infiltrating lymphocytes, or TILs. However, not all TILs are effective at recognizing and attacking tumor cells. "Only a fraction is in fact tumor reactive - the majority are bystanders," Harari explained. "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."

To do this, Harari and his team developed a new AI-driven predictive model, called TRTpred, that can rank T cell receptors (TCRs) based on their tumor reactivity. To develop TRTpred, they used 235 TCRs gathered from patients with metastatic melanoma, already classified as either tumor-reactive or non-reactive. The team loaded the global gene-expression-or transcriptomic - profiles of the T cells carrying each TCR into a machine learning model to identify patterns that differentiate tumor-reactive T cells from inactive counterparts.

"TRTpred can learn from one T cell population and create a rule which can then be applied to a new population," Harari explained. "So, when faced with a new TCR, the model can read its transcriptomic profile and predict whether it is tumor reactive or not."

The TRTpred model analyzed TILs from 42 patients with melanoma and gastrointestinal, lung and breast cancer and identified tumor-reactive TCRs with about 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 that bind strongly to tumor antigens.

"TRTpred is exclusively a predictor of whether a TCR is tumor reactive or not," Harari explained. "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 demonstrated that T cells flagged by TRTpred and the secondary algorithm as both tumor-reactive and having high avidity were more often found embedded within tumors rather 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 diverse tumor antigens. "What we want is to maximize the chances the TILs will target as many different antigens as possible," Harari said.

This final filter organizes TCRs into groups based on similar physical and chemical characteristics. The researchers hypothesized that TCRs in each cluster recognize the same antigen. “So, we pick within each cluster one TCR to amplify, so that we maximize the chances of distinct antigen targets,” said Vincent Zoete, a computational scientist at Ludwig Lausanne who developed the TCR avidity and the TCR clustering algorithms.

The researchers call the combination of TRTpred and the algorithmic filters MixTRTpred.

To validate their approach, Harari's team cultivated human tumors in mice, extracted TCRs from their TILs and used the MixTRTpred system to identify T cells that were tumor-reactive, had high avidity and targeted multiple tumor antigens. They then engineered T cells from the mice to express those TCRs and showed that these cells could eliminate tumors when transferred into the mice.

"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," said Ludwig Lausanne Director George Coukos, a co-author of the study who is planning to launch a Phase I clinical trial that will test the technology in patients.

"Our joint efforts will bring forth a completely new type of T cell therapy."

Pétremand R, Chiffelle J, Bobisse S, Perez MAS, Schmidt J, Arnaud M, Barras D, Lozano-Rabella M, Genolet R, Sauvage C, Saugy D, Michel A, Huguenin-Bergenat AL, Capt C, Moore JS, De Vito C, Labidi-Galy SI, Kandalaft LE, Dangaj Laniti D, Bassani-Sternberg M, Oliveira G, Wu CJ, Coukos G, Zoete V, Harari A.
Identification of clinically relevant T cell receptors for personalized T cell therapy using combinatorial algorithms.
Nat Biotechnol. 2024 May 7. doi: 10.1038/s41587-024-02232-0

Most Popular Now

Philips Foundation 2024 Annual Report: E…

Marking its tenth anniversary, Philips Foundation released its 2024 Annual Report, highlighting a year in which the Philips Foundation helped provide access to quality healthcare for 46.5 million people around...

New AI Transforms Radiology with Speed, …

A first-of-its-kind generative AI system, developed in-house at Northwestern Medicine, is revolutionizing radiology - boosting productivity, identifying life-threatening conditions in milliseconds and offering a breakthrough solution to the global radiologist...

Scientists Argue for More FDA Oversight …

An agile, transparent, and ethics-driven oversight system is needed for the U.S. Food and Drug Administration (FDA) to balance innovation with patient safety when it comes to artificial intelligence-driven medical...

New Research Finds Specific Learning Str…

If data used to train artificial intelligence models for medical applications, such as hospitals across the Greater Toronto Area, differs from the real-world data, it could lead to patient harm...

Giving Doctors an AI-Powered Head Start …

Detection of melanoma and a range of other skin diseases will be faster and more accurate with a new artificial intelligence (AI) powered tool that analyses multiple imaging types simultaneously...

AI Agents for Oncology

Clinical decision-making in oncology is challenging and requires the analysis of various data types - from medical imaging and genetic information to patient records and treatment guidelines. To effectively support...

Patients say "Yes..ish" to the…

As artificial intelligence (AI) continues to be integrated in healthcare, a new multinational study involving Aarhus University sheds light on how dental patients really feel about its growing role in...

Brains vs. Bytes: Study Compares Diagnos…

A University of Maine study compared how well artificial intelligence (AI) models and human clinicians handled complex or sensitive medical cases. The study published in the Journal of Health Organization...

'AI Scientist' Suggests Combin…

An 'AI scientist', working in collaboration with human scientists, has found that combinations of cheap and safe drugs - used to treat conditions such as high cholesterol and alcohol dependence...

Start-ups in the Spotlight at MEDICA 202…

17 - 20 November 2025, Düsseldorf, Germany. MEDICA, the leading international trade fair and platform for healthcare innovations, will once again confirm its position as the world's number one hotspot for...