Predicting age by mining electronic medical records with deep learning characterizes differences between...

Area of Science:

• Computational biology
• Genomics
• Biomedical informatics

Background:

• Assessing physiological age is crucial for personalized and preventative healthcare.
• Electronic medical records (EMR) contain extensive patient data, but their predictive power for chronological age is not well-established.

Purpose of the Study:

• To develop and validate a deep learning model using EMR data to predict chronological age.
• To utilize the discrepancy between predicted and chronological age as a proxy for physiological age.
• To identify health indicators and genetic variants associated with aging and age discrepancies.

Main Methods:

• A deep learning model was trained on 377,686 EMR from the Mount Sinai Health System (MSHS), utilizing vital signs and lab tests.
• The model predicted chronological age with a standard deviation error of approximately 7 years.
• Genome-wide association study (GWAS) was performed on SNP array data from 10,000 patients to identify genetic variants associated with aging.

Main Results:

• The deep learning model accurately predicted chronological age, with higher accuracy for younger and older patients.
• Significant health disparities were observed between patients predicted to be older or younger than their chronological age.
• GWAS identified novel genetic variants linked to inflammation, hypertension, lipid metabolism, height, and lifespan, including genes with missense mutations.

Conclusions:

• EMR data, particularly vital signs and lab tests, can be effectively used to predict chronological age and estimate physiological age.
• The deviation from predicted chronological age serves as a valuable indicator of overall health status.
• This approach can uncover novel genetic associations with aging, paving the way for advanced personalized medicine.