AgeXtend provides transparent explanations for its predictions. Credit: Arora, S. et al.
Scientists have developed AgeXtend, an explainable AI platform designed to discover new geroprotectors – compounds that slow ageing1. The platform integrates machine learning to analyse molecules’ bioactivity, predict toxicity and identify potential mechanisms of action.
Researchers at Indraprastha Institute of Information Technology-Delhi in New Delhi screened over 1.1 billion compounds comprising small molecules, phytochemicals, and cellular and microbial metabolites across databases. They validated promising candidates using lifespan assays in yeast and C. elegans.
The platform accurately identified known geroprotectors like metformin and taurine even when they were excluded from the training data. Its predictions were further tested on microbiome-derived metabolites, revealing potential senescence inhibitors that may delay ageing in human fibroblasts. The model highlights mechanisms linked to ageing, such as cellular senescence, mitochondrial dysfunction and loss of proteostasis.
Unlike black-box AI models, AgeXtend provides transparent explanations for its predictions. Its modular design evaluates geroprotective potential, toxicity risks and likely protein targets for each compound, increasing its reliability for drug discovery.
These findings could pave the way for anti-ageing therapies, offering scalable and rapid identification of new candidates, the researchers say.
