Longitudinal fundus imaging and its genome-wide association analysis provide evidence for a human retinal aging clock
Curation statements for this article:-
Curated by eLife
eLife assessment
This paper is an important contribution to the biological aging field using eye image data to create an aging clock of the retina in data from eyePACS with validation in the UK Biobank. The authors provide compelling evidence that the clock correlates with chronological and phenotypic age, predicting mortality independently of chronological age. The work identifies novel genetic loci with a top site located in the ALKAL2 region, which is functionally validated in a Drosophila model.
This article has been Reviewed by the following groups
Listed in
- Evaluated articles (eLife)
- Computational and Systems Biology (eLife)
Abstract
Biological age, distinct from an individual’s chronological age, has been studied extensively through predictive aging clocks. However, these clocks have limited accuracy in short time-scales. Here we trained deep learning models on fundus images from the EyePACS dataset to predict individuals’ chronological age. Our retinal aging clocking, ‘eyeAge’, predicted chronological age more accurately than other aging clocks (mean absolute error of 2.86 and 3.30 years on quality-filtered data from EyePACS and UK Biobank, respectively). Additionally, eyeAge was independent of blood marker-based measures of biological age, maintaining an all-cause mortality hazard ratio of 1.026 even when adjusted for phenotypic age. The individual-specific nature of eyeAge was reinforced via multiple GWAS hits in the UK Biobank cohort. The top GWAS locus was further validated via knockdown of the fly homolog, Alk , which slowed age-related decline in vision in flies. This study demonstrates the potential utility of a retinal aging clock for studying aging and age-related diseases and quantitatively measuring aging on very short time-scales, opening avenues for quick and actionable evaluation of gero-protective therapeutics.
Article activity feed
-
-
Author Response
Reviewer #2 (Public Review):
This paper reports a novel measure of biological age derived from machine-learning analysis of retinal imaging data with chronological age as the criterion measure. The resulting algorithm is impressive. Not only can the retinal image data accurately predict chronological age in the training data and record changes over short time intervals, but it also proves accurate in independent test data and appears to contain information related to mortality risk. In addition, the authors report a GWAS of the new measure.
I would like to see a bit more validation data in the UKB - how does EyeAge relate to (a) tests of visual acuity - e.g. does it explain aging-related differences?
We have extended the supplemental tables and figures (Supplementary table 5 and Figure 3- figure supplement 2) to …
Author Response
Reviewer #2 (Public Review):
This paper reports a novel measure of biological age derived from machine-learning analysis of retinal imaging data with chronological age as the criterion measure. The resulting algorithm is impressive. Not only can the retinal image data accurately predict chronological age in the training data and record changes over short time intervals, but it also proves accurate in independent test data and appears to contain information related to mortality risk. In addition, the authors report a GWAS of the new measure.
I would like to see a bit more validation data in the UKB - how does EyeAge relate to (a) tests of visual acuity - e.g. does it explain aging-related differences?
We have extended the supplemental tables and figures (Supplementary table 5 and Figure 3- figure supplement 2) to show additional adjustments to the hazard ratios using visual acuity.
(b) measures of morbidity and disability - e.g. how is EyeAge Accel associated with at least some of the counts of chronic diseases, self-reported physical limitations, tests of physical performance, measures of fluid intelligence?
We felt that all-cause mortality is the most clear outcome to test against, as other outcomes were not available for all participants or would require domain-specific knowledge to properly incorporate which we felt was out of scope. Given this, we have added this limitation to the discussion:
“This study has several limitations. First, further work will be needed to assess whether eyeAgeAccel is correlated with other important health outcomes and measures.“
But overall, this is a very strong report of an exciting new biomarker of aging. It was unclear to me whether the algorithm to compute the measure would be publicly available. The authors should clarify.
Code for both training and evaluation of eyeAge from fundus images is available by minimally modifying open-source software we previously released under the permissive BSD 3-clause license. We have added the following “Code availability” section to the paper:
“To develop the eyeAge model we used the TensorFlow deep learning framework, available at https://www.tensorflow.org. Code for both training and evaluation of chronological age from fundus images is open-source and freely available as a minor modification (https://gist.github.com/cmclean/a7e01b916f07955b2693112dcd3edb60) of our previously published repository for fundus model training57.”
-
eLife assessment
This paper is an important contribution to the biological aging field using eye image data to create an aging clock of the retina in data from eyePACS with validation in the UK Biobank. The authors provide compelling evidence that the clock correlates with chronological and phenotypic age, predicting mortality independently of chronological age. The work identifies novel genetic loci with a top site located in the ALKAL2 region, which is functionally validated in a Drosophila model.
-
Reviewer #1 (Public Review):
The authors have used eye image data to create an aging clock of the retina in data from eyePACS with validation in the UK Biobank. They show that the clock predicts mortality independently of chronological age and that it is correlated with phenotypic age. Moreover, a GWAS is conducted in the UK Biobank, which identifies novel genetic loci and a top site located in the ALKAL2 region that is functionally validated in a drosophila model. Overall, the study is interesting with sound methodology and is a nice contribution to the field providing a GWAS summary statistic of the eye clock useful for follow-up analyses.
-
Reviewer #2 (Public Review):
This paper reports a novel measure of biological age derived from machine-learning analysis of retinal imaging data with chronological age as the criterion measure. The resulting algorithm is impressive. Not only can the retinal image data accurately predict chronological age in the training data and record changes over short time intervals, but it also proves accurate in independent test data and appears to contain information related to mortality risk. In addition, the authors report a GWAS of the new measure.
I would like to see a bit more validation data in the UKB - how does EyeAge relate to (a) tests of visual acuity - e.g. does it explain aging-related differences? (b) measures of morbidity and disability - e.g. how is EyeAge Accel associated with at least some of the counts of chronic diseases, …
Reviewer #2 (Public Review):
This paper reports a novel measure of biological age derived from machine-learning analysis of retinal imaging data with chronological age as the criterion measure. The resulting algorithm is impressive. Not only can the retinal image data accurately predict chronological age in the training data and record changes over short time intervals, but it also proves accurate in independent test data and appears to contain information related to mortality risk. In addition, the authors report a GWAS of the new measure.
I would like to see a bit more validation data in the UKB - how does EyeAge relate to (a) tests of visual acuity - e.g. does it explain aging-related differences? (b) measures of morbidity and disability - e.g. how is EyeAge Accel associated with at least some of the counts of chronic diseases, self-reported physical limitations, tests of physical performance, measures of fluid intelligence?
But overall, this is a very strong report of an exciting new biomarker of aging. It was unclear to me whether the algorithm to compute the measure would be publicly available. The authors should clarify.
-
