Sapient has published a study detailing the application of its rLC-MS system for non-targeted metabolomics analysis in 26,042 plasma samples.
The study led to the discovery of key metabolic phenotypes (metabotypes) that correlate with common diseases and enabled development of a machine learning (ML)-based metabolic aging clock found to accurately predict accelerated aging in various chronic diseases, with dynamic reversal of aging following definitive therapy.
A subset of samples from Sapient’s DynamiQ biorepository – comprised of more than 62,000 total plasma samples – was selected from 6,935 individuals to represent diverse demographic backgrounds and disease profiles. The samples were analysed by rLC-MS to capture more than 15,000 metabolites and lipids per sample, providing the first deep view into the comprehensive landscape of human small molecule chemistry. These molecules are derived from both endogenous and exogenous sources and therefore reflect dynamic physiological processes and environmental influences, making them important biomarkers to understand and predict clinically relevant physiological states. Across individuals, biological variation was markedly higher than technical variation, indicating good power to discern biological effects.
Using the large-scale dataset generated via rLC-MS, Sapient identified several distinct subpopulations with metabotypes that correlate with heterogeneous disease phenotypes, including strong associations with cardiometabolic disorders. These findings suggest that plasma metabolomics can be used to metabolically define subgroups, predict disease states, and develop molecular diagnostic tools.
The rLC-MS dataset was also used to train an ML model to predict biological age and capture individual differences in rate of aging, in particular exploring how chronic human diseases may shorten lifespan. This metabolic aging clock successfully predicted accelerated aging in chronic human disorders as well as reversal of aging in kidney disease patients following transplantation, demonstrating that metabolomics may better capture dynamic changes in aging than biomedical data or epigenetic markers alone.
“These findings demonstrate the discovery power of the rLC-MS platform to capture a broad and dynamic landscape of chemical variation in human plasma, across a large population,” said Jeramie Watrous, first author on the paper and co-founder and head of analytical R&D at Sapient.
“The robust, large-scale datasets that can be generated with rLC-MS will substantially increase the ability to identify robust small molecule biomarkers, elucidate novel disease mechanisms, and predict biomedically relevant physiological states.”
“To answer our ambitious questions about shared and unique features of human biology at scale, we had to reimagine the entire technology stack, from our rLC-MS to an AI-driven pipeline for automated feature detection, annotation, and quality control,” said Saumya Tiwari, co-first author and co-founder and head of computational R&D operations at Sapient.
“We built AI-powered custom software and a library of over 13,000 chemical standards to transform raw signal into meaningful biology. The result is a dynamic, high-resolution portrait of human individuality shaped by both biology and environment.”
“We’ve shown that non-targeted metabolomics data can be very valuable to train predictive models of complex physiological states,” said Tao Long, co-founder and head of data science at Sapient.
“With our ML-based analyses, we find circulating metabolites hold close association with key human health and disease phenotypes, and can predict and read out complex, dynamic biological processes including biological aging, disease onset, and therapeutic response.”
Jim Cornall is editor of Deeptech Digest and publisher at Ayr Coastal Media. He is an award-winning writer, editor, photographer, broadcaster, designer and author. Contact Jim here.
