CytoReason and Ferring Using AI to Improve IBD Drug Options

 

CytoReason and Ferring Pharmaceuticals announced last week a collaboration that pairs CytoReason’s artificial intelligence (AI)-based computational model of the human body with Ferring’s expertise in inflammatory bowel disease (IBD). The purpose is to develop cell-based disease models to accelerate drug discovery and reduce development costs.

The collaboration leverages CytoReason’s extensive library of both public and proprietary molecular data to gain critical information on the body’s functioning and address IBD at the cellular level. By the end of this collaboration, CytoReason intends to present insights on the top drug targets in order to provide new therapeutic options for IBD patients.

“The collaboration with Ferring was fairly natural,” David Harel, co-founder and CEO of Israel-based startup CytoReason, told BioSpace. “We are developing in silico models for immunotherapy, cancer, and autoimmune diseases (including IBD).

“Ferring (a multinational corporation headquartered in Switzerland) is a global leader in IBD solutions and drugs. Ferring’s expertise in gastroenterology, immunology and translational medicine will allow us to better understand the complex nature of IBD and, in turn, create more accurate models of the disease with our AI technology.”

The Ferring collaboration is designed for target identification, although the AI approach also can be used for indication assignment before beginning clinical trials.

“IBD is one of the most investigated diseases in the world, second perhaps to cancer and, now, COVID-19,” Harel said. “Not only is it widespread, but it has a meaningful chronic impact on patients’ quality of life.”

The benefit for Ferring is speed and accuracy. “The alternative to AI/machine learning is animal modeling,” Harel pointed out. “An animal model approach could take 18 months and $100,000. Our approach can compare hundreds of diseases in one hour.”

Unlike many bioinformatics programs, CytoReason’s models are based solely on human data. That’s the key feature distinguishing CytoReason’s models from those of other organizations. “There’s no animal data,” Harel reiterated. “That’s important when developing a drug for humans.

“Perhaps more interestingly,” he continued, “because we are not competing with the drug makers, we can train our models on all the clinical trials we are working with. Consequently, our models are improving with every new customer so, over time, they are becoming more accurate.”

CytoReason’s disease models are developed using public and proprietary data. “The type of targets that are derived from our platform can be tailored to the specific needs, requests, and capabilities of our partners,” Harel said. When you have a holistic view of the disease, you can tailor your search,” whether that means large molecules, small molecules, or specific cells or pathways.

With the Ferring collaboration, for example, “We aren’t telling Ferring ‘here are the five best targets,’ because they may differ by client, based upon their validation methods, protocols, manufacturing capabilities, distribution system or other factors. CytoReason takes those factors into account. Therefore, Harel said, “Understanding the disease as a whole is critical.”

CytoReason’s AI/machine learning approach to model design is the natural evolution of biological inquiry that exploded with high throughput screening and the human genome project. Amidst the deluge of data, it was easy to lose the structural and functional information for the tissue, he said.

“Now, rather than developing data using brute force, the approach is to perform high throughput screening on the data rather than upon the physical compounds,” Harel said.

This data-based screening enables access to a continually updated knowledge base.

“Every two minutes, a new journal paper is published. That’s way beyond the ability of a person to comprehend,” Harel pointed out.

CytoReason, however, updates its databases daily to include those new findings. Consequently, an AI/machine learning approach makes it easier to incorporate new knowledge into projects.

Currently, CytoReason has approximately 200 disease models at different levels of resolutions. This flexibility lets researchers examine subpopulations of a disease, thus revealing – oftentimes – that what was considered one disease actually may be multiple, related diseases. Cancer is one example.

“Two diseases may look similar to the physician, but at the molecular level may be very different,” Harel said. That difference helps account for responders and non-responders. “With a robust data model, you can stratify the disease and identify specific elements among disease variants.”

The result is more individualized – and thus more efficacious – therapies.

Industry-wide, companies are challenged by increasing costs and the decreasing probability of success, Harel said. “So, implementing AI is something most companies are pursuing, and it is becoming a key component in drug programs.

“Although we don’t know what drug development will look like 100 years from now, it’s fairly certain to use fewer animals and more technology…more AI. We are helping the industry take the first steps toward that by building in silicomodels for drug discovery and development.”