Deep Learning Blood Test Aging Clock Spans Three Populations

Author: InSilico Medicine
Published: 11 Jan 2018 - Updated: 6 Oct 2026
Publication Details: Peer-Reviewed | Announcement

Table of Contents:
Synopsis - Definition - Overview - FAQs - Insights, Updates - Related Content

Synopsis

This research, published as a peer-reviewed paper in The Journals of Gerontology: Series A, describes an aging clock built by Insilico Medicine that uses deep learning to estimate a person's biological age from ordinary blood test results. The model was trained on a large set of fully anonymized records from Canadian, South Korean and Eastern European patients, and the version trained on all three groups together predicted age more accurately than models built for any single population, which points to real regional differences in how aging shows up in the blood. The clock also proved to be a better predictor of all-cause mortality than chronological age. That matters to seniors, people with disabilities and anyone living with a chronic condition, because a low-cost measure of biological age could let researchers test healthspan treatments in far more people and places, including groups that are often left out of clinical trials.*

At a Glance

• 1 - The clock runs on standard blood test results, so it needs no next-generation sequencing or other specialized equipment.
• 2 - Co-authors came from Gachon University Gil Medical Center, the University of Copenhagen, the University of Alberta and the Biogerontology Research Foundation.
• 3 - Insilico Medicine says it is developing further biomarkers that combine blood biochemistry, transcriptomics and imaging data to track how well interventions perform.
• 4 - Judging longevity interventions by lifespan gains alone would take long and costly longitudinal studies. Reliable aging biomarkers offer a faster way to measure whether a treatment is working.

Topic Definition

Hematological Aging Clock

A hematological aging clock is a computer model that estimates a person's biological age from the results of routine blood tests, such as blood chemistry panels and cell counts. It is trained on large numbers of blood records paired with known ages, so that it learns which patterns in the blood tend to shift as people grow older. When the age it predicts for someone is higher or lower than their actual age, the gap is read as a sign that their body is aging faster or slower than average. Researchers use these clocks as biomarkers of aging, mainly to judge whether treatments intended to extend healthy lifespan are having a measurable effect without waiting decades for survival data.

Overview

Insilico Medicine, Inc., a company specializing in the application of artificial intelligence for drug discovery, biomarker development and aging research, announced in January 2018 the publication of a research paper titled "Population-specific biomarkers of human aging: a big data study using South Korean, Canadian and Eastern-European patient populations" in The Journals of Gerontology: Series A.

In the paper, the authors present a novel deep learning-based hematological human aging clock, a biomarker that predicts the biological age of individual patients. This big data study uses a large dataset of fully anonymized Canadian, South Korean and Eastern European blood test records to train an aging clock.

The developed model predicts the age better than models tailored to the specific populations highlighting the differences of subregion-specific patterns of aging. In addition, the developed clocks were shown to be a better predictor of all-cause mortality than chronological age. The paper includes co-authors from Gachon University Gil Medical Center, University of Copenhagen, University of Alberta, and the Biogerontology Research Foundation.

"If we are to develop actionable biomarkers of aging, we need a comprehensive and robust approach. Such an approach can only be developed using a large number of samples from multiple populations. We are working on multiple biomarkers using deep learning and incorporating blood biochemistry, transcriptomics, and even imaging data to be able to track the effectiveness of the various interventions we are developing," said Polina Mamoshina, a senior research scientist at Insilico Medicine.

"The pursuit of biological aging clocks is a major focus point of the aging field and is a key step in the development of interventions in human aging. This paper represents the evolution of the first easily adaptable clock that can be applied at a population level regardless of population biases. The clock is very cost-effective, without the requirement of next-generation sequencing or other specialized equipment. It is therefore extraordinarily suited for testing aging-interventions in multiple settings across the globe," said Morten Scheibye-Knudsen, MD, Head of the Biology of Aging Laboratory, Center for Healthy Aging, and associate professor, University of Copenhagen.

"Development of effective biomarkers of age is one of the most pressing goals in geroscience today, as it lays the foundation for efficient preclinical and clinical evaluation of potential healthspan-extending interventions. Humans live a long time, and testing the effect of gerontological interventions in humans using lifespan gains as the main criterion for success would be wildly impractical, necessitating long and costly longitudinal studies. By developing accurate biomarkers of aging, the efficacy of potential healthspan-extending interventions could instead be tested according to changes in study participants' biomarkers of age. While significant attention is paid to the development of highly accurate biomarkers of aging, less attention is paid to developing actionable biomarkers of aging that can be tested inexpensively using the tools at hand to the majority of researchers and clinicians. We developed the deep learning-based, blood biochemistry aging clock presented in this paper in the hopes of making progress toward the goal of more actionable biomarkers of aging," said Franco Cortese, co-author of the paper and Deputy Director of the Biogerontology Research Foundation.

"This work demonstrates the synergy between artificial intelligence and aging research. Every living being has age as a feature and it is possible to engage in multi-national collaborations using the very simple data types to assess the population specificity of age predictors. Our group is using advanced AI for multiple clinical applications and has a working collaboration with IBM Watson, but working with Insilico Medicine is a pleasure," said Lee Uhn, PhD, Chief of Artificial Intelligence at the Gachon University Gil Medical Center.

"Age is one of the features possessed by every living creature. In 2015 we made a very neat discovery - when we train the deep neural networks to predict the age of the person, the DNNs capture the most biologically-relevant features and can be re-trained on diseases and can be used to integrate the multiple data types and also extract the most important features within each data type and across the data types. In this paper we show one of the proofs of concept on a very simple and abundant data type that we can now assess the population-specificity of the predictors, the importance of ethnicity and population group in age prediction and the differences in the most important features contributing to the accuracy of these predictors," said Alex Zhavoronkov, PhD, CEO of Insilico Medicine, Inc.

This work may help improve clinical trial enrollment practices, assess the population specificity of a variety of the biomarkers and pave the way for the development of more complex multi-modal biomarkers of aging and disease.

Frequently Asked Questions

NOTE: Researched FAQs by Disabled World (DW)

What is the difference between biological age and chronological age?

Chronological age is the number of years since birth. Biological age is an estimate of how old the body appears to be based on measurable markers such as blood chemistry, and it can be higher or lower than chronological age depending on health, genetics and lifestyle.

Which blood markers are commonly used in blood-based aging clocks?

Blood-based aging clocks typically draw on routine panel results such as albumin, glucose, urea, hemoglobin, cholesterol and red and white blood cell counts. These are the same measures ordered during an ordinary checkup, which is why the approach is inexpensive.

How is a blood test aging clock different from an epigenetic clock?

An epigenetic clock estimates age from DNA methylation patterns and requires specialized laboratory analysis. A blood test aging clock uses standard clinical chemistry and cell count results, which makes it cheaper and easier to run but generally less precise than the best methylation-based clocks.

How accurate are deep learning aging clocks based on blood tests?

Published blood-based deep learning clocks generally estimate age to within roughly five to six years on average. That margin means they are better suited to comparing groups in research than to giving one person an exact age.

Why do aging biomarkers differ between populations?

Blood values that change with age are shaped by genetics, diet, environment, common health conditions and access to medical care, all of which vary by region. Differences in laboratory methods and reference ranges can also shift results between countries.

What does all-cause mortality mean?

All-cause mortality is death from any cause during a set period of time. Researchers use it as a broad outcome measure because it does not depend on correctly assigning a specific cause of death.

Can lifestyle changes lower biological age?

Many of the blood markers used in aging clocks respond to exercise, diet, sleep, smoking status and the management of chronic conditions. Whether moving those markers actually extends lifespan is still being studied, so a lower score should not be read as proof of a longer life.

Can a blood-based biological age score diagnose a disease?

No. A biological age score is a statistical estimate and not a diagnostic test. An unusually high result may be a reason to talk with a doctor, but it cannot identify a specific illness on its own.

Insights, Analysis, and Developments

Editorial Note: An aging clock is only as good as the people it was built on, and that is the quiet lesson in this work. A model tuned to one country's blood values can misread someone from another, which has practical consequences for how volunteers are screened and enrolled in trials and for whether the results of those trials hold up elsewhere. The researchers treat this blood chemistry clock as a proof of concept and a step toward more complex biomarkers of aging and disease, not a finished clinical tool, and the announcement comes from the company that developed it, so independent replication in other populations will be the real test. Even so, a method that works with the routine lab results most clinics already collect is far more likely to reach older adults and people with long-term health conditions than one that depends on costly specialist testing.*

Attribution/Source(s): This peer reviewed publication was selected for publishing by the editors of Disabled World (DW) due to its relevance to the disability community. Originally authored by InSilico Medicine and published on 11 Jan 2018, this content may have been edited for style, clarity, or brevity.

* Editorial additions by Ian C. Langtree.