How Astrophysics Is Powering Cancer Innovation

Concr is all about enabling truly personalized cancer care. Despite all the discoveries and progress made to date, only in a handful of cases can oncologists accurately predict how an individual patient will respond to cancer treatment. We set out to develop an AI tool that could make that prediction [response to cancer therapy] for each cancer patient based on their unique biology, to guide the best treatment and improve outcomes. 

FarrSight® uses advanced Bayesian methods, originally developed with astrophysicists to study the universe, to deeply understand cancer. These powerful algorithms, designed to handle complex, multiscale data – like galaxy signals – allow us to model cancer biology with precision, even with messy and incomplete data.

By adapting these algorithms to oncology, we were able to connect different types of research, molecular and clinical data, building a foundation model of cancer biology. We apply this model to simulate an individual patient’s cancer biology by constructing their molecular digital twin (or FarrSight®-Twin) and use it to predict how that individual will respond to various approved cancer therapies, even when the data is limited.

Most drugs fail in clinical trials because they don’t outperform the standard-of-care (SOC) treatment. By accurately predicting how individual patients will respond to SOC using their clinical and molecular data, FarrSight®-Twin allows us to simulate trial scenarios with precision, as early as Phase 1 clinical trials.

Importantly, by forecasting each participant’s response to SOC treatment, we enable combination trials with SOC in earlier lines of therapy by providing an “n of one” comparator through a digital control. This means better and more informative trial designs, stronger statistical power and a higher probability of drug success with generous time and cost savings. 

Fundamentally, we still don’t understand the intricate nature of biology. Cancer is not a uniform disease; it is different in every individual. We often see biomarker-positive individuals not responding to targeted therapy, while biomarker-negative patients show surprising benefit. So, to unpick these complexities, we need sophisticated tools and approaches that move beyond stratification.

That’s where AI approaches, like Concr’s, can be transformative. The ability to handle uncertainty and complexity, continuous learning and utility over time – these features of our cancer-specific foundation model will help not only with treatment personalization but also with uncovering novel biological mechanisms. But, of course, clinical and scientific communities are key to adoption, so only through collaboration can we advance personalized oncology.

Enabling the selection of the best treatment for every individual based on their molecular biology would impact not only the patient’s outcome, but hopefully their quality of life. With so many good and approved drugs out there, this goal is already attainable and we have studies ongoing to validate that.

Monitoring disease onset and progression is the next level, enabling a truly holistic view of every patient’s journey. This would transform how cancer care is delivered, with the patient firmly back in focus. Yet the impact stretches beyond the individual. Personalized treatment would reduce unnecessary costs, hospital readmissions and ineffective care, delivering tangible savings for healthcare systems. On a broader scale, better outcomes mean more people stay healthy and productive, with a measurable economic impact.