- Home >
- Institut Curie News >
- Immunotherapy: Combining data to better predict efficacy
How to improve the prediction of response to immunotherapy in non-small cell lung cancer? Researchers from Institut Curie, Inserm, and Mines Paris-PSL have taken up this challenge by combining different types of examination data (genomic, radiomic, anatomopathological, clinical) within novel artificial intelligence algorithms. The first article, which has just been published in the Nature Communications journal.
In the vast majority of lung cancers (more precisely in non-small cell lung cancers 1), immunotherapy is prescribed as first-line treatment for 85% of patients. However, some respond to the treatment, while others do not. Thus, successfully predicting its efficacy represents a crucial issue in order to save time on the progression of the disease, avoid unnecessary side effects. and reduce costs. Scientists from Institut Curie, Inserm, and Mines Paris-PSL have embarked on a pioneering project, which is funded by the Arc Foundation and PR[AI]RIE , and which aims to search for new predictive biomarkers.
A pioneer, first of all, in terms of organization: Sixteen researchers from Institut Curie, Inserm, and Mines Paris-PSL, helped by many colleagues and from various fields (imaging, artificial intelligence, pathology, radiomics, tumor biology, etc.) have collaborated in a transdisciplinary manner around the same datasets.
A pioneer in terms of results: this team has managed to identify the best combination of data to predict the response to immunotherapy in non-small cell lung cancer.
Evidence of the benefit of multimodality
"In collaboration with the team of Prof. Nicolas Girard, head of the department of medical oncology at Institut Curie, we collected, for 317 patients, transcriptomic data, i.e. genome expression; radiomic data, i.e. imaging; tumor anatomical pathology data; and finally clinical data," explains Dr Emmanuel Barillot, director of the Computational Oncology Unit (U1331, Institut Curie, Inserm). "We have thus discovered that algorithms that combine data from three or four of these modalities always predict the response to treatment better than those using only one or two. This evidence of the benefit of multimodality had not yet been reported for non-small cell lung cancer."
Even better, scientists have identified the most predictive modalities and linked them to biological mechanisms. "For example, we have observed that the transcriptome provides good quality information, in particular because it makes it possible to quantify dendritic cells—whose action in the response to immunotherapy is already known," elaborates the researcher.
Hope for clinical application in the near future
Discoveries that will have an impact in the short but also the long term. "Our next research will focus on integrating even more data into our algorithms to verify the reliability of predictions and further improve it," announces Nicolas Captier, first author of the study and PhD student in the Computational Systems Biology of Cancer team at Institut Curie. "And ultimately, the hope is to be able to use such algorithms to develop the therapeutic strategy."
The practice will require close collaboration with doctors for its implementation: a process that should be facilitated by the ability of researchers at Institut Curie to work together with teams from the Hospital Group across the board.
[1] These cancers represent more than 80% of lung cancers and include adenocarcinomas (60% of cases), squamous cell carcinomas (30% of cases), and large cell carcinomas (rarer).
[2] Cancer Research Foundation
[3] One of the four French artificial intelligence institutes that brings together PSL University, including Institut Curie, as well as Paris Cité University, the CNRS, Inria, Pasteur Institute, and major industrial players such as Google and Meta.
Reference: Reference: N. Captier et al., Integration of clinical, pathological, radiological, and transcriptomic data improves prediction for first-line immunotherapy outcome in metastatic non-small cell lung cancer, Nature Communications. January 12, 2025. DOI: 10.1038/s41467-025-55847-5
Research News

Breast cancer
Triple-negative breast cancer: new data in the regulation of the mTOR pathway
26/12/2024