This project aims to develop an advanced AI-driven Mixture of Experts (MoE) model to predict health outcomes and multi-morbidity risk, leveraging diverse ageing clocks and biomarkers. The primary research question is: Can integrating data from various ageing clocks, biomarkers, and predictors improve the accuracy of such a predictor, and will this accuracy be sufficient to provide valuable, actionable insight to users? Our objectives include training the MoE model using both publicly available and UK Biobank data, benchmarking it in the Biomarkers of Aging Challenge, and subsequently extending, refining, and deploying it in Rejuve.AI’s Longevity app for real-world application. The scientific rationale is rooted in the need for actionable ageing clocks that predict multi-morbidity, as current models primarily focus on chronological age. By integrating diverse data sources, such as those present in the UK Biobank dataset, data from our partners, as well as data that we will collect through the Longevity app, we aim to create a robust model that not only predicts risk but also engages users in a feedback loop, enhancing model accuracy and contributing to longevity research. This contribution will include not only the predictor’s design, but also the trove of data that the users will collect and share with us in order to receive the predictor’s multi-morbidity risk estimates.
