Breast cancer affects around 48,000 people in France every year, with approximately 12,000 deaths. Our unique AI capabilities place us at the forefront of medical research in the fight against breast cancer.
The Race AI project, an Owkin/Gustave Roussy collaboration, funded by Région Île de France, is one of Owkin’s initiatives to develop solutions for breast cancer treatment. In the application of machine learning to digital pathology slides and clinical data, our study paves the way in breast cancer research by accurately predicting which patients with estrogen receptor (ER)-positive, HER2-negative early breast cancer will relapse.
To learn more about the impact of machine learning on breast cancer patients and oncologists, we spoke to Professor Fabrice André, Head of Research at Gustave Roussy, the leading cancer research institute in Europe.
Owkin: What are the biggest unmet needs in breast cancer?
Fabrice André: There are several important unmet needs in patients with breast cancer. The first is, of course, to cure more patients. Metastatic breast cancer patients that are responsive to endocrine therapy have positive outcomes. So a key unmet medical need is to reverse resistance to endocrine therapy.
The second important strategy is to improve the efficacy of immunotherapy. And the third is to develop complex drugs such as antibodies, drug conjugates, or bi-specific antibodies to improve the outcome of patients with breast cancer.
The final need, that is becoming more and more important, is to avoid toxicity. There are 2 million women around the world, diagnosed with breast cancer every year. Most of them receive toxic therapy. We need to be able to decrease this toxicity with new, less toxic therapies.
Owkin: What does the future of precision medicine look like for breast cancer?
Fabrice André: The future of precision medicine in breast cancer is a world in which clinicians have a holistic view of their patients. Meaning, they fully understand the molecular mechanism of their patient’s cancer evolution. Such as, what is the mechanism of immunosuppression? And, what are the genetic, protein, or metabolic alterations in the livers of their patients that explain toxicity and/or lack of efficacy? Realizing precision medicine relies on the integration of multiple layers of biology to drive a more holistic treatment approach. By firstly, identifying the mechanism of cancer progression and how the patient herself will tolerate the drug, the clinician can schedule the most effective therapeutic strategy.
Owkin: What is the role of machine learning in breast cancer?
Fabrice André: There are many roles now for machine learning. Machine learning in the short term is going to help physicians to interpret image analysis. This can be applied to radiology such as mammograms or scans, but it can also be applied to digital pathology to better predict the outcome of the patient and the sensitivity of the treatment. In the mid-term, machine learning is going to help clinicians to better personalize the therapy, for example, radiation therapy. Then in the long term, machine learning could help us better design drugs, and also model the biology of the patient at the individual level.
Owkin: Can you explain the Race AI project in a few sentences and why you think it is game-changing?
Fabrice André: Race AI is a project done in partnership between Gustave Roussy and Owkin, supported by the Région Île-de-France. The aim of this project is to use digital pathology to predict which Breast Cancer patients have an increased risk of relapse.
Our first results show that this is possible, thanks to machine learning. When we integrate clinical variables and machine learning-based predictors, we can predict relapse from digital pathology slides with an AUC of 0.81. Why is this important? Because if we can predict which patients are going to relapse, we will be able to give this patient new therapy and even investigational therapies in the context of a clinical trial.
Owkin: How does Race AI impact patients and oncologists?
Fabrice André: Race AI will have a significant impact on patients and oncologists. Firstly, this tool could improve the outcome of patients either by decreasing toxic therapy when it is not necessary or by increasing it for high-risk patients to give more aggressive therapy to those that need it. Secondly, Race AI will improve research infrastructure. It’s one of the pioneer projects where hospitals and startup companies have learned to work together to make AI projects applied to cancer. And finally, Race AI will be incorporated into the patient care pathway. So far, pathologists can pull diagnoses from pathology slides. Now, if we have an AI relapse predictor, pathologists will not only be able to read the pathology slides, but they will also have the AI infrastructure to fully interpret the images.
“Race AI is one of the pioneer projects where hospitals and startup companies have learned how to work together to deliver AI projects applied to cancer.”
Owkin: Why is Owkin uniquely positioned to accelerate breast cancer research?
Fabrice André: Owkin is in the unique position to accelerate cancer research for many reasons. Firstly, since Owkin was created by oncologists with a strong culture in translational or applied research in cancer, it is able to move faster than other companies that lack this culture of medical application. Secondly, Owkin has amongst the best data scientists in the world who can make a significant impact in healthcare. And finally, because Owkin has developed the infrastructure and knowledge that specifically addresses the biggest unmet medical needs in cancer research.
“Owkin has amongst the best data scientists in the world who can make a significant impact in healthcare. Additionally, Owkin has developed the infrastructure and the knowledge that specifically addresses the biggest unmet medical needs in cancer research.”
Owkin: What has your experience been working with Owkin?
Fabrice André: I really enjoy my experience working with Owkin. Everyone is kind and pleasant to work with and Owkin offers a good complement to our expertise. In Gustave Roussy, we focus on clinical expertise to identify unmet medical needs and we also have strong expertise in experimental biology. What we need to complement our expertise is skills in data analysis, interpretation, and AI. It’s always pleasant to work with teams that allow us to complement one another. Without major overlaps, each team brings new knowledge to the other. On top of that, our Race AI is a successful project, so overall, it’s the best scenario for us.