Biology is rarely straightforward. As scientists advance in decoding and modifying genes for disease treatment, for example, an increasing collection of data indicates that the proteins and metabolites surrounding those genes cannot be overlooked.
The MIT offshoot ReviveMed has devised a platform for quantifying metabolites — metabolic products like lipids, cholesterol, glucose, and carbohydrates — on a large scale. The firm is utilizing these measurements to unveil why some patients benefit from treatments while others do not, along with gaining a deeper comprehension of disease drivers.
“Traditionally, we’ve managed to measure several hundred metabolites with high precision, but that represents only a small fraction of the metabolites present in our bodies,” states ReviveMed CEO Leila Pirhaji PhD ’16, who co-founded the company with Professor Ernest Fraenkel. “There’s an enormous disparity between what we are accurately quantifying and what actually exists in our system, and that’s the challenge we aim to address. We intend to harness the potent insights from underutilized metabolite data.”
ReviveMed’s advancements occur as the wider medical community is increasingly correlating dysregulated metabolites with illnesses like cancer, Alzheimer’s, and cardiovascular disorders. ReviveMed is employing its platform to assist some of the world’s largest pharmaceutical firms in identifying patients who may benefit from their treatments. It is also providing software to academic researchers at no cost, facilitating insights from previously untapped metabolite data.
“With the rise of AI, we are optimistic that we can address data challenges that have constrained metabolite research,” Pirhaji remarks. “There isn’t a foundational model for metabolomics yet, but we observe how such models are transforming various fields like genomics, so we are initiating their development.”
Identifying a challenge
Pirhaji was born and raised in Iran before arriving at MIT in 2010 to pursue her PhD in biological engineering. She had previously reviewed Fraenkel’s scholarly articles and was eager to contribute to the network models he was creating, which synthesized data from various sources, including genomes, proteomes, and other molecules.
“We were contemplating the broad picture regarding what could be accomplished when everything can be measured — the genes, RNA, proteins, and small molecules such as metabolites and lipids,” explains Fraenkel, who currently sits on ReviveMed’s board. “We’re likely only able to measure around 0.1 percent of small molecules in the body. We believed there must be a method to gain as comprehensive a perspective on those molecules as we have for others. This would enable us to chart out all alterations happening within the cell, whether in the context of cancer, development, or degenerative diseases.”
About midway through her PhD, Pirhaji dispatched some samples to a collaborator at Harvard University for data collection on the metabolome — the small molecules generated by metabolic processes. The collaborator returned a massive Excel sheet filled with thousands of data lines — but they advised her to disregard everything beyond the first 100 rows, as they couldn’t interpret the rest. She viewed that as a challenge.
“I began to ponder if we could leverage our network models to address this issue,” remembers Pirhaji. “There was considerable uncertainty in the data, which intrigued me since no one had attempted this before. It appeared to be a significant gap in the field.”
Pirhaji constructed an extensive knowledge graph encompassing millions of interactions between proteins and metabolites. The data was rich yet chaotic — Pirhaji termed it a “hair ball” that provided no insights about disease. To enhance its utility, she devised a novel approach to characterize metabolic pathways and features. In a 2016 publication in Nature Methods, she detailed the system and utilized it to assess metabolic alterations in a model of Huntington’s disease.
Initially, Pirhaji had no plans to establish a company, but in the concluding years of her PhD, she began to recognize the commercial prospects of the technology.
“There’s no culture of entrepreneurship in Iran,” Pirhaji explains. “I was unaware of how to initiate a business or transform scientific research into a startup, so I took advantage of all that MIT had to offer.”
Pirhaji started enrolling in courses at the MIT Sloan School of Management, including Course 15.371 (Innovation Teams), where she collaborated with peers to explore applications for her technology. She also utilized the MIT Venture Mentoring Service and MIT Sandbox, and participated in the Martin Trust Center for MIT Entrepreneurship’s delta v startup accelerator.
When Pirhaji and Fraenkel formally established ReviveMed, they collaborated with MIT’s Technology Licensing Office to access the patents connected to their work. Pirhaji has since further refined the platform to address additional challenges she identified from discussions with numerous leaders in pharmaceutical firms.
ReviveMed commenced its work with hospitals to investigate how lipids are disrupted in a condition known as metabolic dysfunction-associated steatohepatitis. In 2020, ReviveMed partnered with Bristol Myers Squibb to anticipate how subsets of cancer patients would respond to the company’s immunotherapies.
Since then, ReviveMed has collaborated with numerous companies, including four of the top 10 global pharmaceutical firms, assisting them in comprehending the metabolic mechanisms within their treatments. These insights facilitate the quicker identification of patients likely to gain the most from various therapies.
“If we can determine which patients will benefit from each drug, it would significantly reduce the complexity and duration of clinical trials,” Pirhaji states. “Patients will receive appropriate treatments more swiftly.”
Generative models for metabolomics
Earlier this year, ReviveMed compiled a dataset derived from 20,000 patient blood samples, which it utilized to create digital twins of patients and generative AI models for metabolomic research. ReviveMed is making its generative models accessible to nonprofit academic researchers, potentially accelerating our understanding of how metabolites impact various diseases.
“We’re advancing the democratization of metabolomic data use,” declares Pirhaji. “It’s infeasible for us to access data from every single patient globally, but our digital twins can be employed to identify patients who might benefit from treatments based on their demographics, for example, by detecting patients who could be predisposed to cardiovascular disease.”
This endeavor forms part of ReviveMed’s mission to develop metabolic foundation models that researchers and pharmaceutical firms can employ to comprehend how diseases and treatments alter patients’ metabolites.
“Leila resolved numerous challenging issues encountered when attempting to advance an idea from the lab to something robust and reproducible enough for biomedicine application,” Fraenkel observes. “In the process, she also recognized that the software she developed is incredibly powerful on its own and has the potential to be transformative.”