AI Predicts Biological Age Better Than DNA Tests Using Simple Health Reports
Large language models outperformed telomere length and epigenetic clocks in predicting aging and mortality risk across 10+ million people.
Summary
Researchers developed an AI system using large language models to predict biological age from routine health examination reports. Testing across six population studies with over 10 million participants, the AI outperformed traditional aging measures like telomere length, frailty scores, and epigenetic clocks in predicting mortality risk. The system achieved superior accuracy for both overall aging and organ-specific aging assessment, offering a cost-effective alternative to expensive genetic tests for large-scale health monitoring.
Detailed Summary
Accurate biological age assessment is crucial for identifying health risks and preventing age-related diseases, but current methods face significant limitations in cost, accessibility, and predictive power.
Researchers from Tsinghua University developed a novel framework using large language models (LLMs) to estimate biological age from standard health examination reports. They validated this approach across six population-based cohorts encompassing over 10 million participants.
The LLM-predicted biological age achieved a concordance index of 0.757 for all-cause mortality prediction, significantly outperforming traditional aging proxies including telomere length, frailty index, eight epigenetic age measures, and four machine-learning models. Each year of accelerated aging (age gap) was associated with a 5.5% increased mortality risk.
The system also demonstrated superior performance in organ-specific aging assessment, better predicting corresponding organ diseases compared to conventional machine-learning approaches. The researchers successfully applied age gaps to identify proteomic biomarkers of accelerated aging and developed risk prediction models for 270 diseases.
This breakthrough offers a precise, reliable, and cost-effective approach for aging assessment that could revolutionize personalized health management in large populations, making biological age testing accessible through routine medical examinations rather than expensive specialized tests.
Key Findings
- LLM biological age prediction outperformed telomere length and epigenetic clocks for mortality risk
- Each year of accelerated aging increased all-cause mortality risk by 5.5%
- System successfully predicted organ-specific aging and disease risks
- Framework identified proteomic biomarkers and predicted risk for 270 diseases
- Method works with standard health reports, making it cost-effective for large populations
Methodology
Large-scale validation study across six population-based cohorts with over 10 million participants. LLM framework trained to predict biological age from routine health examination reports, compared against established aging measures including telomere length, frailty index, and epigenetic clocks.
Study Limitations
Study based on abstract only - full methodology details unavailable. Validation limited to specific population cohorts which may affect generalizability. Long-term clinical outcomes and intervention effectiveness using this framework require further investigation.
Enjoyed this summary?
Get the latest longevity research delivered to your inbox every week.