Cell Metab. 2026 Apr 21. pii: S1550-4131(26)00108-7. [Epub ahead of print]
Jiawei Chen,
Ya Ren,
Yong Zhou,
Ziyang Wang,
Kehang Mao,
Zhengqing Yu,
Jiyang Li,
Xiaoxiao Guo,
Hao Xu,
Yiyang Wang,
Yi Wang,
Bo Pang,
Hongxiao Liu,
Huiru Tang,
Jing-Dong J Han.
Understanding aging and complex diseases requires diverse data, ranging from molecular profiles to imaging and routine clinical tests. However, most multi-omic datasets measure only a subset of modalities and are confounded by batch effects. Here, we present AURORA (AI unification and reconstruction of omics reassembly atlas), a generative deep-learning platform that integrates seven modalities (including transcriptomics, metabolomics, microbiome, 3D and thermal facial imaging, and clinical laboratory tests) across 581,763 samples from 425,258 individuals. AURORA harmonizes batch effects and reconstructs missing data across modalities, enabling highly accurate multimodal aging clocks and disease risk predictors. It also supports personalized in silico perturbation analyses to predict intervention and drug responses, validated using longitudinal cohorts. As a proof of concept, we provide a prototype AI agent that converts single-input modalities into a multimodal report for users and researchers. Together, AURORA links non-invasive inputs to comprehensive aging biomarkers and therapeutic discovery.
Keywords: aging clocks; biological aging; digital twin; disease risk prediction; drug repurposing; facial imaging; generative AI; in silico perturbation; multi-omics integration; personalized medicine