A world first. The year 2022 ended under the best auspices for the Curie Institute. In partnership with the Israeli start-up Ibex, the teams at the Paris center have published just before Christmas a study in a journal of the group Nature highlighting an artificial intelligence (AI) capable of identifying very precisely the type of breast cancer from which a patient suffers, from the simple biopsy of her tumor. An unprecedented performance in the context of a clinical study, which demonstrates the usefulness of AI to facilitate the work of pathologists and prevent them from missing an anomaly.
AI is showing more than promising results in the fight against cancer. This year again, it will have the honors of the great congress of the American Society of Clinical Oncology, which will be held from June 2 to 6 in Chicago. Scientists are developing tools capable of modeling a tumor and analyzing its characteristics, so as to adapt treatments as closely as possible to the patient’s needs. With an approach that no longer classifies cancers according to the organ of origin, but according to their biological characteristics. Mathematicians, biologists and oncologists are joining forces to advance AI and develop these “digital twins” of tumors. We are seeing the multiplication of unprecedented collaborations such as the one announced in mid-May by the Gustave-Roussy cancer center, the CentraleSupélec engineering school and Inserm. These three institutions co-founded with the University of Paris-Saclay and Unicancer the PRISM university hospital institute dedicated to precision medicine. Concretely, about thirty research teams will meet within this consortium which aims to become “the largest campus in Europe dedicated to cancer”.
Their approach to “deep cohort” [NDLR : des cohortes de patients étudiés de façon très détaillée], as promising as it is complex, consists of analyzing a colossal amount of data from the sequencing of the tumor genome, but also proteins, RNA, the immune system or the patient’s microbiota. “We can measure tens of thousands of pieces of information per cell”, specifies Paul-Henry Cournède, director of research at CentraleSupélec.
The dream of personalized medicine
Just as Airbus creates digital duplicates of its planes or EDF of its nuclear power plants, the PRISM center models tumors from this vast data set, runs simulations and obtains predictions that will be closer and closer to reality as algorithms train on increasing numbers of patients. “We build a digital twin and we observe how it behaves to assess, for example, the effectiveness of such a drug on such a type of tumour”, summarizes Paul-Henry Cournède. With the line of sight, the dream of personalized medicine.
This meticulous work of analysis must indeed lead to “the creation of a digital map summarizing the biology of each patient”, explains the Gustave-Roussy center. It will identify new biomarkers and predict the body’s reaction to drugs. “For a whole range of cancers, we already know how to identify known mutations and deduce the right treatment”, underlines THE Professor Fabrice André, director of research at the centre.
AI is also proving valuable for analyzing tumor organoids, a sort of scale model of cancer, built from tumor cells that researchers culture and replicate up to hundreds of times. They then test various drugs and dosages on these mini-tumors. “Via image analysis, AI makes it possible to follow precisely how the organoid responds to treatment”, explains Paul-Henry Cournède. The start-up Orakl, co-founded by Fanny Jaulin, research director at Gustave-Roussy, s specializes in this technology of the future. Founded in Marchshe plans to raise funds this summer.
France has assets to promote
The public authorities want to develop a “culture of collaboration” between researchers, doctors and data scientists. This is the ambition of the Paris-Saclay Cancer Cluster. THE hub will host investors, incubate biotech and multiply partnerships with pharmaceutical groups eager to find the next breakthrough innovation. “The mechanics are launched and should quickly gain momentum,” said Paul-Henry Cournède.
It will certainly take time to catch up with the most prolific ecosystems in the world such as Boston or Cambridge. But France has some assets to put forward. The Curie and Gustave-Roussy centers welcome internationally renowned researchers. With 13 Fields medals, France dominates world mathematics. Grandes écoles like Polytechnique, Télécom Paris or CentraleSupélec excel in applied mathematics. Under these favorable conditions, France has given birth to some promising nuggets in AI against cancer, such as Tribun, which uses image analysis algorithms to improve diagnoses, Owkin, which applies the IA to anatomopathology data, or from TheraPanacea, which develops digital models to maximize the effectiveness of radiotherapy.
Faced with a disease that remains in France the leading cause of death in humans and the second for women, the government wants to accelerate the movement. Endowed with an envelope of 33 million euros and supported by the France 2030 program, the PortrAIt consortium aims to create “at least 15 AI tools for digital pathology” within five years. “Thanks to these funds, the industrial sector will structure itself more quickly, with the obligation to have an impact on the patient”, appreciates Fabrice André. “Publish studies in Nature Or Science, it’s very good, but the most important thing is to help the sick to get better”, adds Paul-Henry Cournède.
Digital twins also have another virtue, less spectacular but just as crucial to fight cancer effectively: they allow patients to visualize their disease. To this end, Dassault Systèmes is working on a tumor visualization model. “Over 30% of patients are non-adherent to their treatment. Understanding their cancer better will have a positive impact on treatment adherence [NDLR : la capacité à prendre correctement son traitement]“, predicts Fabrice André.
Finally, the AI has another interest: the algorithms will help doctors to distinguish at an increasingly early stage the “common” cancers from the most difficult to treat. The former can be taken care of in “classic” hospitals and clinics when the latter require going to advanced centers such as Curie or Gustave-Roussy. A better distribution of patients will increase efficiency in the face of a disease which remains the scourge of our century.