A novel study to calculate immune-aging from peripheral blood T lymphocyte subsets and their mitochondrial parameters in healthy Chinese subjects.

BACKGROUND

Immunosenescence is a process in which the body's immune function declines with age, which is associated with the increased risk of infection, tumors, and other diseases. Traditional biological age assessment is difficult to fully reflect the changes in immune function. Flow cytometry can obtain multi-dimensional data such as immune cell subsets and mitochondrial function, and combined with machine learning, it is possible to quantify immune-aging. This study aimed to establish an accurate immune-aging prediction model based on peripheral blood indicators of healthy people in China.

METHODS

Peripheral blood samples were collected from healthy people aged 0.5 to 89 years old from September 2023 to December 2023, and 72 indicators (including blood routine, biochemistry, T/B/NK cell subsets, mitochondrial mass, and percentage of low membrane potential, etc.) were detected. A total of 11 core features were obtained by multi-stage screening of automatic and manual feature construction and mutual information-Boruta-forward selection. Random forest, LightGBM, and other algorithms were used, combined with the Bagging fusion strategy for modeling, and cancer patients were used as the external validation set.

RESULTS

The LightGBM model had the best performance, with an R² of 0.809 and an MAE of 5.687 in the test set of the healthy population. The immune-aging of cancer patients was significantly higher than the actual age. The model had a strong discrimination (AUC>0.900) in the adolescent group (0-19 years old) and the elderly group (≥60 years old), and a slightly weaker discrimination (AUC>0.780) in the middle-aged group.

CONCLUSION

This model can accurately quantify the degree of immune senescence, provide reference for health management, vaccination, and anti-aging intervention, and also provide new ideas for the judgment of aging-related diseases.