Philadelphia healthcare nonprofit wields AI to find new uses for old drugs
The Philadelphia healthcare nonprofit is developing a knowledge graph to predict which existing treatments could have wider application A Philadelphia-based nonprofit, Every Cure, uses AI to identify missed opportunities in drug repurposing, a process that involves finding new uses for existing medicines to treat diseases that were not initially intended or developed. The use of these methods is referred to as "drug repurpose," as there is no financial incentive for pharmaceutical companies to pursue this research. Every Cure uses multiple algorithms to compare all drugs and all diseases, making it unique in its approach. The company has also secured $48.3 million over three years from the Advanced Research Projects Agency for Health (ARPA-H), a federal agency established in 2022 to support biomedical and health research. It plans to hire data scientists and data engineers next year to double its team to 30, with the goal of having a publicly available knowledge graph by 2024.
Published : a month ago by Sarah Huffman in Business Health Tech
Grant Mitchell thinks most drugs on the market aren’t being used to their full potential, so he launched a company to fix that.
Called Every Cure, the Philly-based nonprofit identifies missed opportunities — and helps connect more people to treatments that work. Finding ways to use existing medicines to treat diseases for which they weren’t initially intended or developed is known as “drug repurposing.”
Every Cure is using AI to make drug repurposing faster and more effective.
“Many diseases have shared mechanisms and pathologies,” Mitchell, who acts as company CEO, told Technical.ly. “One drug can work for multiple diseases.”
The problem is that 80% of drugs on the market are generic, he said, so there’s no financial incentive for pharmaceutical companies to pursue the kind of research needed to make those connections.
He and cofounders Tracey Sikora and David Fajgenbaum incorporated the nonprofit in 2022. Fajgenbaum’s interest in the situation was personal. The University of Pennsylvania professor had developed Castleman Disease, a lymph node condition that at the time had no approved treatments. Yet looking at existing drugs, they found a 25-year old medicine to treat his disease.
Every Cure uses multiple algorithms to compare all drugs and all diseases, which is what makes it unique, Mitchell said.
While other organizations doing this kind of work may focus on a specific drug or disease, Every Cure looks at all cases, with the goal of making the connections that will impact the greatest number of patients.
“[We] can really zoom in on what are the opportunities that are most likely to work in the clinic,” Mitchell said. “Then filter that for the ones that can help patients the most, and that we can validate the quickest.”
Repurposing can be faster than traditional drug development process because the drug has already been tested for safety in humans, according to the National Center for Biotechnology Information. This means drugs can get to patients more quickly, skipping to the preclinical or clinical trial stage.
Every Cure is building a biomedical knowledge graph, which documents known relationships between biomedical concepts such as drugs, genes, proteins, cell types, tissue types and organ systems. The organization uses large language models to review medical literature and identify relationships between biomedical concepts.
The company was a winner this February in the Cure Xchange Challenge: Health AI for Good, scoring up to $1 million in funding, space on Cure’s New York City life sciences campus, mentoring and networking opportunities.
Every Cure also recently secured $48.3 million over three years from the Advanced Research Projects Agency for Health (ARPA-H), a federal agency established in 2022 to support biomedical and health research.
Over the next year, Every Cure plans to hire data scientists and data engineers, doubling its team to 30, with the goal of having a publicly available knowledge graph to share with researchers by the end of 2024.
Beyond predicting connections between drugs and diseases, the founders hope Every Cure will be able to test the predictions in labs and eventually help patients through clinical trials, Mitchell said.
Even just a few years ago, he noted, it wasn’t possible to review all medical literature and identify connections the way AI and large language models now can.
“The ability to make inferences between drugs and disease to these complex pathways and interconnected networks — it’s really a perfect example of a question and a problem better suited for a machine than a human,” Mitchell said. “We simply couldn’t comprehend or integrate this level of data all at once.”
Topics: Nonprofits, AI