Crunching fruit fly and human data via machine learning reveals key metabolites that impact lifespan in both species

Study suggests that threonine may hold promise as therapeutic target for aging

Discoveries that impact lifespan and healthspan in fruit flies are usually tested in mice before being considered potentially relevant in humans, a process that is expensive and time intensive.  A pioneering approach taken at the Buck Institute leapfrogs over that standard methodology.

Utilizing cutting-edge machine learning and systems biology, researchers analyzed and correlated huge data sets from flies and humans to identify key metabolites that impact lifespan in both species.  Results published online in Nature Communications suggest that one of the metabolites, threonine, may hold promise as a potential therapeutic for aging interventions.

 “These results would not have been possible without this pioneering approach,” says Buck professor Pankaj Kapahi, PhD, senior author of the paper.  “There is a lot of data sitting out there that is not being correlated between species. I think this approach could be a game-changer when it comes to identifying potential interventions to improve human health.”

Threonine has been shown to protect against diabetes in mice. The essential amino acid plays an important role in collagen and elastin production and is also involved in blood clotting, fat metabolism and immune function.  

The method – simplified

The work began with former Buck postdoc Tyler Hilsabeck, PhD, crunching data (involving metabolomics, phenotypes and genomics) to analyze 120 metabolites in 160 strains of fruit flies on both restricted and normal diets. The goal was to reveal how different genotypes responded to the diets to influence lifespan and healthspan. “This allowed us to find the ‘needles in the haystack’ when it came to identifying relevant metabolites,” Hilsabeck says. 

Vikram Narayan, PhD, a postdoctoral fellow then cross-referenced findings with human data from the massive UK Biobank. “Using the human data allowed us to focus on interesting metabolites to those that are conserved in both species. It also allowed us to uncover the impact of those metabolites in humans,” he says.  Importantly, the team then brought those relevant metabolites back into the fly to validate results.

The results

In flies, threonine extended lifespan in a strain-and-sex-specific manner. Individuals with higher levels of threonine-related metabolites had longer, healthier lives.  “We’re not saying that threonine is going to work in all conditions,” says Kapahi. “Our research shows it works in subsets of both flies and people.  I think most of us have stopped expecting to find a ‘magic-bullet’ intervention for aging.  Our method provides another way to develop precision medicine for geroscience.” 

The results also include findings that were not so positive for both species.  Orotate, which is relatively understudied and has been linked with fat metabolism, was negatively associated with aging. In flies orotate counteracted the positive impact of dietary restriction across every strain of the animals.  In humans, orotate was linked to a shorter lifespan. 

Larger implications

Kapahi hopes the larger research community will begin employing this method. “So many times we find things that work in worms and flies and then we don’t have the resources to move the basic science forward. This approach allows us to say with a lot more certainty that discoveries are going to be relevant in humans.” Kapahi says this method may reduce the need for studies in mice, something he welcomes.

CITATION: Systems biology and machine learning approaches identify metabolites that influence dietary lifespan and healthspan responses across flies and humans

DOI: https://doi.org/10.1038/s41467-024-52909-y

Other Buck researchers involved in the study include Vikram P. Narayan, Kenneth A. Wilson, Enrique Carrera, and Judith Campisi.  Other contributors include Daniel Rafferty, Northwest Metabolomics Research Center, University of Washington, Seattle; Daniel Promislow, Departments of Pathology and Biology, University of Washington, Seattle; and Rachel B. Brem, Department of Plant and Microbial Biology, University of California, Berkeley.

Acknowledgements: This work was supported in part through funds from the National Institutes of Health, F31AG062112, T32AG000266-24, F31 AG052299, T32AG000266-23, R56AG038688, R21AG054121, RO1AG045835 and AG049494, as well as the Hevolution Foundation, American Federation of Aging Research and the Larry L. Hillblom Foundation.