CAMBRIDGE, Mass.: #AI--ReviveMed, an MIT spinout at the forefront of artificial intelligence and precision medicine, today announced the release of a new preprint on bioRxiv describing mzLearn, a novel, data-driven metabolite signal detection algorithm. mzLearn automates signal detection without requiring any parameters or prior knowledge and corrects for instrument drift, overcoming critical limitations in data quality and reproducibility that have hindered large-scale metabolomic studies. This innovation powers ReviveMed’s first-in-class, pre-trained generative AI models, opening new frontiers in metabolomics research.

Generative AI has transformed fields such as natural language processing, genomics, and single-cell transcriptomics. Now, ReviveMed is extending these groundbreaking capabilities to metabolomics, an area rich with biomarker and therapeutic discovery opportunities yet largely unexplored. This milestone is particularly timely, as GLP-1 drugs highlight the potential of metabolic interventions to reduce cancer, diabetes, and cardiovascular disease risks. ReviveMed’s platform is uniquely positioned to make groundbreaking contributions to these fields.

“Generative AI has revolutionized other areas of biomedicine, and with mzLearn, we are unlocking its transformative potential for metabolomics,” said Dr. Leila Pirhaji, CEO and Co-Founder of ReviveMed. “Our platform not only detects high-quality metabolite signals at an unprecedented scale but also captures critical metabolic variations across populations, enabling more precise patient stratification and deeper insights into disease biology.”

In extensive evaluations using over 20,000 blood-based metabolomics profiles from diverse cohorts, mzLearn demonstrated robust metabolite signal detection, laying the foundation for ReviveMed’s pioneering pre-trained generative models. These models effectively captured metabolite representations associated with demographic and clinical variables, leading to superior clinical predictions. Notably, they outperformed established clinical-grade risk scores in predicting renal cell carcinoma (RCC) patient outcomes. These results highlight the transformative potential of ReviveMed’s data-driven approach and generative models to deliver earlier and more precise disease diagnosis, enhanced prognostic stratification, and refined therapy-response predictions.

The preprint, titled “mzLearn, a data-driven LC/MS signal detection algorithm, enables pre-trained generative models for untargeted metabolomics,” is now publicly available on bioRxiv. mzLearn is accessible to non-profit academic researchers at mzLearn.com, democratizing untargeted metabolomics data and paving the way for developing foundation models for metabolomics.

About ReviveMed
ReviveMed is an MIT spinout dedicated to harnessing large-scale metabolomics data and cutting-edge machine learning to advance precision medicine. By unlocking the full potential of metabolomic insights, ReviveMed drives innovation across clinical development and fuels the discovery of life-changing treatments.

Fonte: Business Wire