NEW YORK – Gero, a biotechnology startup that is tackling human aging and longevity with the help of artificial intelligence, took an unusual step this month by announcing the "commencement of discussions" to collaborate with publicly traded Foxo Technologies on longitudinal epigenetic biomarker discovery.
The companies have not yet signed a letter of intent, but "a deal will happen," Gero CEO Peter Fedichev said. The scientific, technical, and financial details are still being worked out.
Fedichev said that he has known Foxo leadership for several years and the two companies have been "exchanging ideas" for a while since epigenetic biomarker developer Foxo has a deep interest in longevity.
The firms are now coming together first to validate some of Gero's earlier work, with the hope that a collaboration could eventually inform clinical studies, according to Foxo CSO Brian Chen.
Fedichev and two Gero colleagues published a preprint on BioRxiv last year that explored the notion of a "thermodynamic biological age" that they based on DNA methylation signatures and longitudinal electronic medical records gleaned from UK Biobank.
"This is cooperation with the goal to better understand the dynamics of human aging [with] one of the most variable sources of data, which is DNA methylation," Fedichev said.
The epigenetic data Gero used to this point has been cross-sectional. The company was attracted to Foxo because like UK Biobank, the latter has a collection of longitudinal DNA methylation data, according to Fedichev.
"They have some biological insights that they've discovered through other datasets, and they would like to first apply that to our longitudinal datasets," Chen explained.
Foxo has both long-term and shorter-term longitudinal data, from a variety of sources, the largest of which is the Physicians' Health Study, a large-scale trial initiated by Brigham and Women's Hospital in 1982, according to Chen.
"That allows us and Gero to look at how resilient or how persistent or how stable certain measurements are over time, especially in relation to some sort of … anti-aging intervention or any other lifestyle interventions," he explained. "And that gives us insights as to whether we could actually change these measures or if there's an inherent state that's very stable."
Fedichev, who has a Ph.D. in theoretical physics from the University of Amsterdam, said that unlike physics, biology often neglects trajectories.
Fedichev said that Gero's goal is to conduct "high-quality, radical science in the field of longevity and [provide] value" to researchers of chronic diseases.
"The mission of Gero is a substantial increase in average and maximum human health span and lifespan," added Alex Kadet, the firm's chief business officer. "We believe that with our models, we can delineate effects of aging from the effects of diseases, and for the first time, enable data-driven, unbiased target discovery in chronic diseases."
Fedichev said that trajectory prediction calls for "modern machine learning and lots of data to start doing time-resolved biology."
To train its AI, Gero uses genomic and clinical data from UK Biobank, SomaLogic, and its customers. Electronic medical records are the foundation of its analysis because they are longitudinal.
"We used these EMRs to build a generative model of human health so we can predict future diseases from [a cohort's] current diseases," Kadet said.
"Our goal is to develop therapeutics against aging and age-related diseases," Kadet said. Any discovery will be done in partnership with established pharmaceutical companies.
To this end, Gero in January formed a research collaboration with Pfizer to apply machine learning for fibrotic disease-related target discovery. The startup received an unspecified upfront payment and can earn additional milestone payments in the multiyear partnership.
Gero is using clinical records and exome sequences that Pfizer has collected to, among other things, understand why patients who are predisposed for fibrosis may not have the condition. "When you have trajectories, you know what is causing what," Fedichev said.
Kadet said that this type of research is more effective than genome-wide association studies alone because GWAS will "regress out" biological age and potentially lead investigators down dead ends.
"If you run a GWAS for chronic kidney disease or something like that, you [may] end up with the same targets for diabetes or atherosclerosis, and you're lost," Kadet explained. "But when we use our models, it helps to separate" phenotypes, he added.
Fedichev explained that it is important to corroborate phenotype-based predictions with longitudinal DNA methylation data, which is exactly what Minneapolis-based Foxo offers. "Foxo is a very natural partner to do that because they have been collecting DNA methylation samples," he said.
Fedichev said that Gero's models have not previously been trained on longitudinal DNA methylation data.
Fedichev said that the collaboration will be more investigational than with a clear commercial goal. While there has been some longitudinal human epigenetics research with cohorts of identical twins, Gero sees the addition of clinical data as largely uncharted territory.
Gero was founded in 2018 in Russia and is legally headquartered in Singapore, though operations are mostly in the US now. Fedichev said that the firm will likely soon be moving all of its assets to Palo Alto, California.
Kadet said that the company had been planning on moving to Silicon Valley before Russia was globally sanctioned in the wake of its invasion of Ukraine in February 2022. "We feel that if you do biotech and especially longevity biotech, the valley is the place to be," he said.
Gero has raised more than $7.5 million to date, most recently in a $2.2 million Series A round that closed in 2020. The firm expects to announce another round in about a month.
The timing of the funding announcement is somewhat dependent on the Foxo deal getting done. Chen laughed as he called himself the "bottleneck" in creating a formal collaboration as he fleshes out details on the Foxo side.
Once an agreement is in place, the technology integration should only take a few days. Chen said Foxo would just have to create a workspace in its Amazon Web Services cloud environment and authorize Gero to access his company's data.
Chen said that the integration would start with a proprietary Foxo dataset and later branch out to datasets from some of the firm's other partners.
"The point is to look at very broad and very diverse datasets to see how robust this finding is and how it might change in response to an intervention," Chen explained. "[Many] researchers try to go for cross-sectional data biobanks, but that doesn't tell you anything about how an aging signature or molecular signature changes over time."