Anagnostakis Filippos, Ko Sarah, Saadatinia Mehrshad, Wang Jingyue, Davatzikos Christos, Wen Junhao
Laboratory of AI and Biomedical Science (LABS), Columbia University, New York, NY, USA.
Artificial Intelligence in Biomedical Imaging Laboratory (AIBIL), Center for AI and Data Science for Integrated Diagnostics (AI2D), Perelman School of Medicine, University of Pennsylvania, Philadelphia, PA, USA.
Nat Commun. 2025 May 26;16(1):4871. doi: 10.1038/s41467-025-59964-z.
Multi-organ biological aging clocks across different organ systems have been shown to predict human disease and mortality. Here, we extend this multi-organ framework to plasma metabolomics, developing five organ-specific metabolome-based biological age gaps (MetBAGs) using 107 plasma non-derivatized metabolites from 274,247 UK Biobank participants. Our age prediction models achieve a mean absolute error of approximately 6 years (0.25<r < 0.42). Crucially, including composite metabolites (e.g. sums or ratios of raw metabolites) results in poor generalizability to independent test data due to multicollinearity. Genome-wide associations identify 405 MetBAG-locus pairs (P < 5 × 10/5). Using SBayesS, we estimate the SNP-based heritability (0.09< < 0.18), negative selection signatures (-0.93 < S < -0.76), and polygenicity (0.001<Pi < 0.003) for the 5 MetBAGs. Genetic correlation and Mendelian randomization analyses reveal potential causal links between the 5 MetBAGs and cardiometabolic conditions (e.g., metabolic disorders and hypertension). Integrating multi-organ and multi-omics features improves disease category and mortality predictions. The 5 MetBAGs extend existing biological aging clocks to study human aging and disease across multiple biological scales. All results are publicly available at https://labs-laboratory.com/medicine/ .
不同器官系统的多器官生物衰老时钟已被证明可预测人类疾病和死亡率。在此，我们将这个多器官框架扩展到血浆代谢组学，利用来自274247名英国生物银行参与者的107种血浆非衍生化代谢物，开发了五个基于器官特异性代谢组的生物年龄差距（MetBAGs）。我们的年龄预测模型的平均绝对误差约为6岁（0.25<r<0.42）。至关重要的是，由于多重共线性，纳入复合代谢物（例如原始代谢物的总和或比率）会导致对独立测试数据的泛化性较差。全基因组关联确定了405个MetBAG-基因座对（P<5×10⁻⁵）。使用SBayesS，我们估计了5个MetBAGs的基于单核苷酸多态性的遗传力（0.09<<0.18）、负选择特征（-0.93<S<-0.76）和多基因性（0.001<Pi<0.003）。遗传相关性和孟德尔随机化分析揭示了5个MetBAGs与心脏代谢疾病（例如代谢紊乱和高血压）之间的潜在因果联系。整合多器官和多组学特征可改善疾病分类和死亡率预测。这5个MetBAGs扩展了现有的生物衰老时钟，以在多个生物尺度上研究人类衰老和疾病。所有结果均可在https://labs-laboratory.com/medicine/ 上公开获取。
