A study published in iScience Journal, in collaboration between scientists from China, the UK and the USA, presented a proof of concept for an aging clock based on microbiomic data. Deep Neural Networks (DNN) were identified as the best performing algorithm for age prediction, achieving a mean absolute error (MAE) of 10.60 years in cross-validation and 5.91 years in independent testing. This accuracy surpasses the baseline median age assignment. The study found that Elastic Net (EN) was unsuitable for intestinal age prediction.
A notable observation is that all models predicted higher ages for younger donors with Type 1 Diabetes (T1D), suggesting possible microbiome alterations associated with the disease. The researchers hypothesize that diabetes may accelerate aging through gut microbiome mechanisms.
The aging clock’s primary function is to measure biological time, but it also helps to understand the aging process. Using Accumulated Local Effects (ALE), the study identified gut microbes linked to aging. However, the results of ALE vary based on dataset structure and model implementation.
Certain gut microbes, like Bacteroides and Bifidobacterium, showed significant effects on age prediction. The study also noted the importance of Akkermansia muciniphila, associated with healthy aging, and butyrate-producing bacteria, which might influence aging.
The presence of certain poorly described bacteria in the human gut, observed in the study, may be due to batch effects or could provide new insights into intestinal aging. The study suggests further research with larger datasets and considering various variables.
The findings demonstrate that microbiome profiles can accurately predict age in healthy individuals and indicate accelerated aging in T1D patients. This could lead to the development of new aging biomarkers and applications like intestinal rejuvenation through dietary interventions, pending a deeper understanding of gut community dynamics and metabolism.
Photo Copyright © 2020 Fedor Galkin et al. Open access article under the CC BY-NC-ND license.