Introduction
The advent of whole genome sequencing has significantly improved our knowledges of role of genes in pathogenesis, yet genes themselves only account less than 10% of the diseases.1 The exposome, on the other hand, play a bigger role in biological aging and the development of systemic diseases processes.2,3 Specifically, the exposome refers to the integrated compilation of all physical, chemical, biological, and psychosocial influences—such as air pollution, microplastics, nutrition, psychosocial stress, socioeconomic status, and physical activity—that impact biology.4-6 To capture the complexity of the exposome, researchers collect and analyze a wide range of biosamples, including serum, urine, saliva, hair, and other biological matrices, to monitor environmental exposures and their biological effects. While blood samples (serum or plasma) are widely used and particularly valuable for assessing persistent chemicals and circulating metabolites,7 they are less suitable for non-persistent compounds with short half-lives, where concentrations in blood may fall below detection limits with high-resolution mass spectrometers-based non-targeted analysis. In such cases, urine or hair can provide more appropriate matrices for capturing exposure profiles.8 Advances in mass spectrometry have expanded the ability to detect and annotate both exogenous chemicals and endogenous metabolites across these matrices, though the majority of data are semi-quantitative rather than strictly quantitative.9
In recent years, precision environmental health (PEH) has emerged as a new discipline that explores the complex interplay between the exposome and biological systems across the human life course.10 Building upon the exposome framework, PEH integrates exposomic data with genomic and molecular information for more personalized health insights. Omics-based biomarkers, including genomics, epigenomics, transcriptomics, proteomics, metabolomics, lipidomics, and microbiomics, provide critical insights into human physiology and pathophysiology, elucidating the molecular mechanisms through which environmental exposures influence human health and contribute to disease development.11 Various omics-based biological clocks, such as epigenetic and proteomic clocks, have been proposed, offering tools to quantify biological aging.12,13 While current multi-omics technologies offer powerful tools for deciphering molecular changes, these are often costly, complex and many are invasive, thus limiting the ability to scale the application of these technologies across larger populations in community settings.
Multi-modal ocular biomarkers approach may serve as a novel, cost-effective, and promising complement to the current omics-based biomarkers in PEH. National Institute of Health (NIH) BEST (Biomarkers, EndpointS, and other Tools) resource categorizes biomarkers into different purposes, including diagnostic, response, monitoring, and predictive functions, highlighting their versatility in addressing diverse clinical and research needs.14 Though ocular biomarkers have been conventionally used for diagnosing ocular diseases and monitoring treatment response, there is emerging evidence that they can be used to monitor the impact of the exposome on systemic health and predict systemic disease risk over time.15-19 Based on synthesis of the emerging evidence, we proposes future directions that integrates ocular biomarker into next generation of PEH research.
Ocular imaging biomarkers
The human eye holds great potential for biomarker discovery due to its unique anatomy and accessibility for non-invasive imaging. (Figure 1) Color fundus photography (CFP) captures high-resolution images of the retina, enabling the assessment of structural biomarkers such as vessel caliber and vessel tortuosity, which reflect vascular health. Optical coherence tomography (OCT) provides cross-sectional and three-dimensional views of the retina, enabling detailed evaluation of specific biomarkers such as retinal layer thickness, the presence of pathological fluid (eg, edema or subretinal fluid).20 Optical coherence tomography angiography (OCTA) delivers functional insights into retinal and choroidal blood flow without requiring dye injections, making it particularly valuable for detecting microvascular abnormalities, such as changes in capillary density and foveal avascular zone (FAZ) area.21 Advances in artificial intelligence (AI) have enabled automated extraction of quantitative ocular biomarkers with remarkable accuracy and repeatability, making them a scalable and clinically relevant tool for large-scale population health studies.22-24
Illustration of various biomarkers based on ocular anatomy. CFP = Color fundus photography; OCT = Optical coherence tomography; OCT-A = Optical coherence tomography angiography; AH = Aqueous humor; VH = Vitreous humor. Created with BioRender.com.
Imaging biomarker for environmental exposures
Retinal imaging biomarkers are intricately connected to environmental exposures and psychosocial factors, providing valuable insights into how these exposures contribute to microvascular and neurodegenerative changes.25-27 Genome-wide association studies (GWAS) suggest that genetic factors account for less than 20% of the variations in retinal vasculature, underscoring the substantial role of environmental and lifestyle factors in shaping retinal microvascular features.28 Table 1 summarizes current studies on common environmental exposures and their associated ocular imaging biomarker changes. Notably, similar alterations—such as decreased central retinal artery equivalent (CRAE), increased central retinal vein equivalent (CRVE), and thinning of the retinal nerve fiber layer (RNFL) and ganglion cell inner plexiform layer (GCIPL)—have been observed across diverse exposures, including particulate matter (pm), smoking, and psychosocial stressors. This convergence of findings suggests shared biological pathways through which distinct exposures exert overlapping vascular and neurodegenerative effects.
Summary of current studies on environmental exposures and ocular biomarkers.
| Environmental exposure | Ocular biomarker | Reference |
|---|---|---|
| Imaging biomarker | ||
| Particulate matter | Decreased CRAE; increased CRVE; RNFL and GCIPL thinning | 25,26,101-103 |
| Smoking | Decreased CRAE, DCP vessel density and CVI; increased CRVE; RNFL thinning | 104-108 |
| Dietary pattern | High carbohydrate food and low fiber intake is associated with decreased CRAE and increased CRVE; heart-healthy diet is associated with decreased tortuosity and increased FD | 32,109,110 |
| Psychosocial stress | Increased CRVE; decreased CRAE | 27 |
| Socioeconomic status | Low SES was associated with low AVR and GCIPL thinning | 111,112 |
| Physical activity | Decreased CRVE; increased CRAE; increased RNFL; smaller FAZ | 33,113-115 |
| Molecular biomarker | ||
| Particulate matter | Increased cytokine in tears | 56,116,117 |
| Smoking | Decreased tear secretion and increase acute phase reactant proteins; decreased Ascorbic Acid in AH; increased VEGF-A in VH | 116,118,119 |
| Psychosocial stress | Increased cortisol in tears | 120 |
| Physical activity | Decreased oxidative stress marker | 121 |
Abbreviations: AH, aqueous humor; AVR, arteriolar-to-venular ratio; CRAE, central retinal artery equivalent; CRVE, central retinal vein equivalent; CVI, choroid vascularity index; DCP, deep retinal capillary plexus; FAZ, foveal avascular zone; FD, fractal dimensions; GCIPL, ganglion cell inner plexiform layer; RNFL, retinal nerve fiber layer; VH, vitreous humor; VEGF, vascular endothelial growth factor.
Mechanistically, chronic exposure to environmental and psychosocial stressors can induce vascular changes through mechanisms like endothelial dysfunction and inflammation. Stress-related hormonal responses impair endothelial nitric oxide (NO) synthesis, a key regulator of vascular tone, and generate reactive oxygen species, which degrade NO and disrupt vascular homeostasis.29,30 Inflammation further mediates these effects, with elevated pro-inflammatory cytokines (eg, IL-6, TNF-α) and C-reactive protein being strongly associated with retinal venular widening, even in children.30,31 These vascular changes may also be indirectly exacerbated by stress-related lifestyle factors such as reduced physical activity and poor dietary habits.32,33 Taken together, these mechanisms provide a biological basis for the observed imaging signatures, reinforcing the value of retinal biomarkers as sensitive indicators of systemic processes—including inflammation, oxidative stress, and microvascular dysfunction.
Imaging biomarker for systemic health
Retinal biomarkers not only reflect environmental exposures but are also closely linked to systemic diseases.28,34 In 2020, the term oculomics was introduced to describe the study of the associations between ocular biomarkers and systemic health conditions, positioning the retina as an emerging critical component in precision health initiatives.35 Table 2 summarizes current studies on ocular imaging biomarkers and systemic health outcomes. Notably, the same biomarkers implicated in environmental exposures—such as increased CRVE, decreased CRAE, and thinning of the RNFL and GCIPL—are also observed across chronic diseases including hypertension, hyperlipidemia, chronic kidney disease (CKD), coronary heart disease, and Alzheimer’s disease and related dementias (ADRD). Phenome-wide association studies have correlated retinal biomarkers with multiple clinical phenotypes and the Eye Biomarker Database consolidates 889 biomarkers across 26 ocular diseases and 939 biomarkers associated with 181 systemic diseases.36 Recent advances in AI have enabled the prediction of multiple systemic disease from retinal images with remarkable accuracy.19,37-40
Summary of current studies on ocular imaging biomarkers and systemic health outcome.
| Systemic diseases | Ocular biomarker | Reference |
|---|---|---|
| HTN | Increased CRVE; decreased CRAE, tortuosity, SCP and DCP vessel density; choroidal thinning | 122-126 |
| HLD | Increased CRVE; decreased CRAE | 106,127 |
| T2DM | Decreased SCP vessel density; Increased FAZ; RNFL and GCIPL thinning | 128,129 |
| CHD | Increased CRVE; decreased CRAE and FD | 130 |
| CKD | Increased CRVE; decreased CRAE and SCP vessel density | 131,132 |
| Stroke | Increased CRVE; decreased CRAE and FD | 130,133 |
| ADRD | Increased CRVE; decreased CRAE and FD; RNFL and GCIPL thinning; increased FAZ; choroidal thinning; increased CDR | 44,134-136 |
Abbreviations: CRAE, central retinal artery equivalent; CDR, cup to disc ratio; CRVE, central retinal vein equivalent; DCP, deep retinal capillary plexus; FAZ, foveal avascular zone; FD, fractal dimensions; GCIPL, ganglion cell inner plexiform layer; SCP, superficial retinal capillary plexus; RNFL, retinal nerve fiber layer.
The retina’s diagnostic potential lies in its unique ability to integrate and reflect systemic processes due to its shared embryological origin, vascular architecture, and neurovascular connections with other organs. As part of the microcirculation, the retinal vasculature mirrors systemic endothelial health, responding to circulating inflammatory mediators, oxidative stress, atherosclerosis, and hemodynamic changes.41,42 These processes drive vascular remodeling, including arteriolar narrowing, venular widening, and altered vessel tortuosity, while also contributing to neurodegenerative changes, such as amyloid-beta accumulation and tauopathy seen in ADRD.43,44 These pathways are central to the “common soil hypothesis,” implicating shared mechanisms—such as inflammation, oxidative stress, and vascular dysfunction—in multiple chronic diseases, including cardiometabolic diseases, neurodegenerative disorders, and CKD.45 These interconnected processes influence retinal and choroidal thickness, vascular permeability, and optic nerve structure, further underscoring the retina’s value as a window into systemic health.
Beyond structural parameters, composite aging biomarkers derived from retinal images provide powerful, non-invasive indicators of systemic aging. The retinal age gap—defined as the difference between predicted retinal age from fundus photographs and chronological age—has been consistently associated with all-cause mortality, cause-specific mortality, and multiple systemic morbidities.46-48 Building on this framework, fundus images have also been used to estimate PhenoAge, a validated composite of nine blood-based biomarkers of biological age, giving rise to the retina-derived RetiPhenoAge.49,50 This biomarker has been shown to outperform conventional markers such as telomere length and grip strength in predicting morbidity and mortality, and has demonstrated strong associations with cognitive decline, dementia, cerebral small vessel disease, and proteomic signatures of aging.51 Together, these retinal imaging–based biological age markers complement structural parameters by providing scalable, non-invasive, and mechanistically informative indicators of systemic aging across diverse diseases.
Ocular molecular biomarkers
Beyond imaging biomarkers, the eye’s liquid-filled chambers—tears, aqueous humor (AH), and vitreous humor (VH)—offer a rich reservoir of biomarkers for exposomic analysis. (Figure 1) These fluids are in constant exchange with both the external environment and systemic circulation, thereby capturing signals of external exposures and internal physiological responses. Compared with traditional matrices such as blood, tears can be collected in a minimally invasive manner, making them well suited for temporal monitoring of the exposome.52 AH and VH offer complementary information to blood, providing localized insights into ocular pathophysiology and capturing metabolic or inflammatory changes that may not be detectable systemically.
Molecular biomarker for environmental exposures
Tear, located at the air–tear interface, is particularly sensitive to airborne pollutants.52 Studies have shown that tears capture markers of PMs, volatile organic compounds, trace metals, and nicotine within minutes to hours of exposure, highlighting their value for short-term exposure assessment.52 More broadly, microplastics (MPs) have now been identified across multiple ocular compartments, including tears, AH, and VH, with their presence associated with alterations in key ocular biomarkers such as tear break-up time, Schirmer’s test scores, intraocular pressure, and vitreous opacities.53-55 While analytical methods for micro- and nanoplastic detection are still being refined, converging evidence from these studies suggests that MPs are not confined to the ocular surface but can accumulate throughout different eye matrices, reinforcing the eye’s potential as a unique system for monitoring environmental exposures.
In addition to capturing external exposures, ocular fluids also reflect the body’s internal physiological responses. Tears and AH contain inflammatory and angiogenic mediators that change dynamically with pollutant exposure. Studies have shown significant increases in inflammatory markers like IL-6, IL-8, and vascular endothelial growth factor (VEGF) in tear with increased exposure to air pollution, highlighting its ability to reflect localized and systemic inflammatory effects.56 Experimental and animal studies shows that ocular surface exposure to microplastics triggers oxidative stress, inflammation, reduced goblet cell density, and microbial dysbiosis.57 Beyond these molecular changes, the ocular surface microbiome has emerged as a key internal exposome. Ambient air pollutants disrupt microbial homeostasis, reducing commensal diversity while enriching pathogenic taxa such as Staphylococcus, Bacteroidia, and Klebsiella, which correlate with goblet cell depletion and chronic ocular inflammation.58
Molecular biomarker for systemic health
Systematic evidence now links tear protein and metabolite changes to a wide spectrum of systemic diseases, including neurodegenerative, autoimmune, and metabolic conditions.59 For example, in ADRD, alterations in tear protein composition have been reported, including elevated lactoferrin, which is thought to reflect amyloid-related immune dysregulation and oxidative stress pathways.59 In multiple sclerosis, tear proteomics has revealed increased immunoglobulins and pro-inflammatory cytokines, consistent with systemic immune activation and blood–brain barrier dysfunction mirrored at the ocular surface.59 Similarly, in type 2 diabetes (T2DM), specific metabolites such as carnitine, tyrosine, and uric acid are elevated in tears, and anti-diabetic drugs like metformin can also be detected, illustrating how tear fluid integrates both disease-related metabolic perturbations and treatment exposure.60 Building on this, tear-derived extracellular vesicles (EVs) represent a promising next-generation biomarker platform. EVs protect their protein and RNA cargo from degradation and can cross from the bloodstream into tears, offering a stable and information-rich signal.61 They have shown implicated in multiple autoimmune diseases such as Dry Eye Disease, Sjögren’s, graft-versus-host disease.61
In addition to tear-based biomarkers, AH and VH provide a complementary perspective, enabling deeper molecular understanding of ocular and systemic aging processes through advanced omic technologies. For instance, using a combination of liquid-biopsy proteomics and AI, researchers have developed proteomic clocks based on the AH and VH specimens that can assess cellular aging within non-regenerative tissues, such as the eye, in vivo.62 These tools reveal that diseases like ADRD not only alter molecular profiles but also accelerate cellular aging processes in the retina, highlighting the interplay between systemic diseases and ocular aging.62
Research prospects for future
Apply a systemic approach to study impact of exposome on ocular biomarkers
Despite advances in understanding the role of environmental exposures, the study of their impact on retinal biomarkers remains fragmented. A systematic approach is needed to capture the effects of exposome on retinal structure and function. One promising framework is the Exposome-Wide Association Study (ExWAS).63 Inspired by GWAS, which identify associations between genetic variants and disease outcomes, ExWAS adopts a similar agnostic, high-dimensional design to test hundreds of environmental exposures simultaneously for association with a phenotype. Methodologically, ExWAS typically uses logistic or Cox regression with multiple-testing correction.64 More recent implementations incorporate machine learning techniques such as shrinkage models, deletion/substitution/addition algorithms for multi-exposure modeling.65 Building on ExWAS, Exposome-Wide Interaction Study (ExWIS) extends the framework to model effect modification and biological interactions.66 ExWIS systematically tests for interactions between pairs of exposures, or between exposures and genetic risk scores, thereby capturing non-additive and synergistic effects.
Complementary to single-exposure frameworks such as ExWAS, mixture analysis methods are increasingly being used to evaluate the combined effects of multiple, correlated exposures. Methods such as Bayesian Kernel Machine Regression enable flexible modeling of non-linear and interactive effects within mixtures, while Quantile G-Computation provides an interpretable summary estimate of the joint impact of a set of exposures.67,68 By moving beyond single-exposure associations, mixture methods provide a more realistic representation of the exposome and open new opportunities for understanding how complex exposure environments, in interaction with social determinants of health, shape disease risk. Applied to ocular biomarkers, these approaches could clarify how social exposome—including education, food insecurity, financial status, healthcare access—contribute jointly to retinal thinning and vascular caliber changes.69
Explore ocular biomarkers as mediators between environmental exposures and systemic health outcome
Mediation analysis is a statistical and epidemiological approach used to understand the mechanisms through which an exposure influences an outcome. Specifically, it partitions the total effect of an exposure into a direct effect (the portion not explained by intermediates) and an indirect effect (the portion that operates through one or more mediators).70 Mediators are typically defined as intermediate variables or biomarkers that lie on the causal pathway between an exposure and an outcome, helping to explain how or why the exposure leads to disease.70 By identifying and quantifying these indirect pathways, mediation analysis provides mechanistic insight that complements traditional association studies and can highlight potential intervention targets.70 One example of this approach in environmental health is the “meet-in-the-middle” framework, where intermediate biomarkers are evaluated based on their dual association with exposures and outcomes, their plausibility in linking the two, and their reproducibility across studies.71
Historically, clinical biomarkers and socioeconomic factors have been used as mediators between environmental exposures and systemic health outcomes, with recent efforts expanding to include multi-omics biomarkers.72,73 The development of novel statistical frameworks has enabled high-dimensional mediation analysis, allowing researchers to simultaneously test multiple potential mediators—an approach particularly well-suited for incorporating omics data.74-77 While multi-omics approaches are valuable for uncovering complex biological pathways, they are often limited by high costs, technical complexity, the need for advanced equipment, and small sample sizes, which reduce statistical power and generalizability. In contrast, retinal biomarkers present a more accessible and practical alternative. As mentioned in earlier paragraphs, changes in retinal biomarkers such as CRAE, CRVE, and RNFL thickness have each been associated with multiple environmental exposures as well as systemic outcomes. The quantitative shifts in these parameters can therefore act as mediators, providing measurable biological pathways through which the exposome impacts human health. For instance, CRAE maybe a mediator between psychosocial stress and cardiovascular disease outcome, while retinal age gap could be a mediator between dietary inflammation index and all cause mortality.46,78,79 (Figure 2) Imaging modalities such as OCT and CFP are cost-effective, minimally invasive, highly reproducible, and widely available in clinical settings, enabling larger sample sizes and more robust mediation analyses.24
Schematic figure showing the association between exposomics, molecular changes, ocular biomarkers, and systemic health outcome in a conceptual mediation framework. Created with BioRender.com.
Integrate multi-modal biomarker for comprehensive mechanistic insight
The crosstalk between structural biomarkers and molecular pathways could reveal the biological underpinnings of observed microstructural changes, offering a window into systemic health dynamics. The IMAGINE study revealed significant correlations between intraretinal fluid features detected through advanced OCT and intraocular cytokine levels, such as VEGF and monocyte chemotactic protein-1 from AH.80,81 To be specific, VEGF levels were associated with increased central subfield thickness and intraretinal fluid volume, highlighting VEGF's role in fluid dynamics and vascular permeability.80,81 This connection to molecular features can potentially help us better understand the molecular mechanisms behind micro-structural changes, thus using eye as model organ to better understanding environmental impact on systemic health. Although the study’s relatively small sample size limits its generalizability, it provides a critical foundation for future research. Larger-scale studies are necessary to validate these findings and expand their applicability.
The rapid advance of foundation models (FMs) presents an exciting opportunity to scale the integration of structural and molecular biomarkers with environmental determinants of health. Unlike conventional models, FMs are trained in a self-supervised manner on massive, heterogeneous datasets and can operate in a multimodal fashion, jointly processing images, omics, clinical text, and geospatial data within a shared representational space.82-84 This flexibility makes them uniquely suited for exposome research, where missing data and diverse modalities are common. A recent study demonstrated this potential by predicting both overall and organ-specific biological age from routine health data, with strong associations to mortality that surpassed classical biomarkers.85 Current retinal imaging clocks have shown that age gaps derived from fundus and OCT images predict mortality and age-related diseases, while proteomic clocks capture cell-type–specific aging processes not visible on imaging.46,62 Integrating these approaches through a multimodal FM could yield a more comprehensive “Multi-Modal Ocular Aging Clock” that links structural changes with molecular signatures and environmental exposures. Proof-of-concept studies such as OmiCLIP, which align omics and imaging via contrastive learning, illustrate how such integration can be achieved, offering a scalable framework for using the eye as a model organ to uncover how the environment shapes biological aging.86
Limitations & opportunities
One of the most pressing challenges in the field of oculomics is the need for standardized “big” data. The power of oculomics studies largely stems from their reliance on vast datasets, often incorporating tens of thousands of retinal images. Similarly, ExWAS usually requires a large sample size to yield statistically significant results. However, the intersection of these two fields presents a significant hurdle: very few datasets currently exist that include both comprehensive exposomics data and detailed retinal imaging on a scale sufficient to conduct robust analyses. Table 3 summarizes current datasets current datasets with exposome and ocular imaging data. Notably, two major large scale cohorts for epidemiological research in the United States, All of Us, and Million Veteran Program currently both do not collect or provide retinal imaging data for research purpose.87,88 UK Biobank is so far the only national biobank that incorporates multi-modal ocular imaging, in an organized and easily retrievable format.89 Its rich modalities of data, comprehensive coverage of diseases, and large sample size have made it an essential part of most of the exposomics and oculomics studies nowadays. A few disease-specific cohorts, such as the Atherosclerosis Risk in Communities (ARIC) study and the Chronic Renal Insufficiency Cohort (CRIC), include ocular imaging data.90,91 However, large exposome cohorts such Human Early Life Exposome (HELIX) and Personalized Environment and Genes Study (PEGS) do not collect retinal imaging as one of their data modalities.64,92 The scarcity of such large, integrated datasets limits the ability to fully explore the potential connections between environmental exposures and retinal biomarkers, which in turn hinders our capacity to uncover novel insights into disease mechanisms. To overcome this limitation, multi-center collaborations and international consortium initiatives will be essential. The National Heart, Lung, and Blood Institute for example convened a workshop in 2022 that emphasized the importance of standardized retinal imaging protocols, interoperable data structures, and scalable analytics to advance the use of retinal biomarkers in cardiovascular and systemic disease research.42
Summary of current cohorts with environmental exposures and ocular imaging data.
| Cohort | Baseline population | Exposure data | Ocular imaging | ||
|---|---|---|---|---|---|
| Environmental sensor | Biospecimen | CFP | OCT | ||
| Large scale national cohort | |||||
| UKBiobank89 | 500 000 individual aged 40-69 years in UK | + | + | + | + |
| All of US88 | 1 million individuals in US | + | + | + | – |
| Million Veteran Program87 | 1 million military veterans in US | – | + | – | – |
| Rotterdam Study137 | 15 000 individuals in Netherlands | + | + | + | + |
| China Kadoorie Biobank (CKB)138 | 512 000 adults aged 30-79 years from China | + | + | – | – |
| Human Phenotype Project139 | 100 000 individuals globally | – | + | + | – |
| PRECISE SG100K140 | 100 000 individuals aged 21-84 in Singapore | – | + | + | + |
| Medium to large cohort on specific diseases | |||||
| Artificial Intelligence Ready and Exploratory Atlas for Diabetes Insights (AI-READI)141 | 4000 individuals with and without diabetes from US | + | + | + | + |
| Atherosclerosis Risk in Communities (ARIC)142 | 15 800 individuals (aged 45–64 years) in US | + | + | + | – |
| Singapore Epidemiology of Eye Disease (SEED)143 | over 10 000 Malay, Indian and Chinese individuals aged over 40 years in Singapore | – | + | + | + |
| Beaver Dam Eye Study144 | 5000 individuals aged 43-84 years from Beaver Dam, Wisconsin | – | + | + | – |
| Blue Mountain Eye Study145 | 3600 individuals aged > 50 years from Australia | – | + | + | – |
| Multi-Ethnic Study of Atherosclerosis146 | 6800 individuals aged 45-84 years from four ethnic groups (White, African American, Hispanic, and Chinese) in US | + | + | + | – |
| Beijing Eye Study147 | 4439 residents aged ≥ 40 years in Beijing, China | – | + | + | + |
| Chronic Renal Insufficiency Cohort (CRIC)148 | 5500 individuals in US | – | + | + | – |
Abbreviations: CFP, Color fundus photography; OCT, Optical coherence tomography.
Another critical challenge is the lack of standardized exposomics profiling. Currently, even if researchers identify several cohorts with rich environmental exposome data, inconsistencies in biometrics, sample collection and storage protocols, and lab analytical platforms can make it difficult to integrate the omics data across cohorts. For instance, the composition of tears has high temporal variability, making it difficult to compare data across different studies.93,94 To overcome these barriers, the establishment of internationally accepted guidelines for data normalization, harmonization, and sharing is urgently needed. The Banbury Exposomics Consortium has made infrastructure development and the establishment of data standards for harmonization one of top priorities for exposome research.95 Beyond biospecimen analysis, questionnaire-based data collection also requires rigorous standardization. Although most cohorts incorporate questionnaires, these are rarely uniform and are often tailored for specific study objectives rather than capturing the full breadth of the exposome. A notable example of efforts to address this challenge is the PEGS, a large and demographically diverse North Carolina-based cohort.64,96 PEGS collects exposome data through multiple surveys, including the Health and Exposure Survey, the Internal Exposome Survey, and the External Exposome Survey, which together capture over 1000 variables on lifestyle, occupational and residential exposures, medication use, diet, sleep, stress, and infectious diseases.64,96
The collection and management of such extensive datasets also raise significant concerns about data privacy, particularly when data are gathered from multiple sites. These exposomics and retinal imaging data are highly sensitive personal information, making it crucial to implement robust data protection measures to ensure confidentiality. The use of AI algorithms to analyze retinal images may introduce biases if the training data is not representative of diverse populations.97 Beyond population diversity, it is equally important to ensure that the data encompasses a wide range of systemic and ocular conditions, as these significantly influence retinal changes. A lack of representation in such conditions could hinder a comprehensive understanding of their effects and interactions. This gap may further exacerbate healthcare disparities, particularly for individuals with limited access to advanced ocular imaging technologies, who may be excluded from the analysis.97 Addressing these ethical and logistical challenges is crucial to ensure that the integration of exposomics and oculomics is both scientifically robust and equitable.
One potential solution to address data privacy concerns is through approaches such as federated learning, a machine learning technique that enables multiple institutions or devices to collaboratively train a shared model without transferring raw data to a central repository.98,99 Instead, the data remains local, and only model updates, such as gradients or weights, are shared and aggregated. This approach can enhance privacy while still enabling comprehensive analysis. In parallel with these privacy-preserving strategies, global collaborative efforts are equally important to tackle the pressing issue of diversity and representation in ocular imaging datasets.
The GlobalRetFound Consortium is one initiative designed to address the lack of diversity in ocular imaging data.100 It aggregates contributions from over 100 countries and combines synthetic data generation with selective real data to build a more representative global dataset. By minimizing the need for direct raw data transfer, the consortium supports data privacy while expanding demographic and geographic diversity. Similar international consortia will be essential to integrate oculomics and exposomics, ensuring that future precision environmental health research is both inclusive and equitable.
Conclusion
Ocular biomarkers hold great potential to advance PEH research because they are non-invasive, scalable, and capable of capturing both structural and molecular signatures of systemic processes. Retinal imaging modalities such as CFP, OCT, and OCTA provide reproducible measures of vascular health, neurodegeneration, and biological aging, while ocular fluids capture molecular signatures of environmental exposures and systemic responses. Together, they offer a cost-effective and accessible complement to traditional omics-based biomarkers. Looking forward, novel research methodologies—including ExWAS, high dimensional mediation analysis, and multimodal foundation models—offer exciting opportunities to better integrate ocular biomarkers with environmental data. Nonetheless, significant challenges remain. The scarcity of large datasets linking exposome and ocular data, lack of standardized profiling methods, and issues of data privacy and equity limit progress. Addressing these gaps through international collaboration, harmonization of data collection, and equitable technology deployment will be crucial. By overcoming these barriers, ocular biomarkers can become a cornerstone of PEH, enabling earlier detection, targeted prevention, and healthier longevity.
Author contributions
Haoran Cheng (Conceptualization [lead], Formal analysis [lead], Methodology [lead], Visualization [lead], Writing—original draft [lead], Writing—review & editing [lead]), Jeremy A. Sarnat (Funding acquisition [equal], Writing—review & editing [equal]), Douglas I. Walker (Funding acquisition [equal], Writing—review & editing [equal]), Anant Madabhushi (Funding acquisition [equal], Writing—review & editing [equal]), Amritpal Singh (Visualization [equal], Writing—review & editing [equal]), Rohan Dhamdhere (Visualization [equal], Writing—review & editing [equal]), Jodhbir S. Mehta (Writing—review & editing [equal]), Tien Yin Wong (Funding acquisition [equal], Writing—review & editing [equal]), John S. Ji (Writing—review & editing [equal]), Carmen J. Marsit (Funding acquisition [equal], Writing—review & editing [equal]), Dean P. Jones (Funding acquisition [equal], Writing—review & editing [equal]), Daniel S. W. Ting (Conceptualization [equal], Funding acquisition [equal], Methodology [equal], Supervision [equal], Writing—review & editing [equal]), Darren S. J. Ting (Conceptualization [equal], Funding acquisition [equal], Methodology [equal], Supervision [equal], Writing—review & editing [equal]), and Donghai Liang (Conceptualization [equal], Funding acquisition [lead], Methodology [equal], Supervision [lead], Writing—review & editing [equal]). All authors approved the final manuscript.
Funding
DL and JAS are supported by the National Institute of Health (NIH) research grant [R01ES035738], and DL, JAS, DIW, DPJ, and CJM are supported by the NIH Center Grant [P30ES019776]. am was supported by the National Cancer Institute under award numbers R01CA268287A1, U01CA269181, R01CA26820701A1, R01CA249992-01A1, R01CA202752-01A1, R01CA208236-01A1, R01CA216579-01A1, R01CA220581-01A1, R01CA257612-01A1, 1U01CA239055-01, 1U01CA248226-01, 1U54CA254566-01, National Heart, Lung and Blood Institute 1R01HL15127701A1, R01HL15807101A1, National Institute of Biomedical Imaging and Bioengineering 1R43EB028736-01, VA Merit Review Award IBX004121A from the United States Department of Veterans Affairs Biomedical Laboratory Research and Development Service the Office of the Assistant Secretary of Defense for Health Affairs, through the Breast Cancer Research Program (W81XWH-19-1-0668), the Prostate Cancer Research Program (W81XWH-20-1-0851), the Lung Cancer Research Program (W81XWH-18-1-0440, W81XWH-20-1-0595), the Peer Reviewed Cancer Research Program (W81XWH-18-1-0404, W81XWH-21-1-0345, W81XWH-21-1-0160), the Kidney Precision Medicine Project (KPMP) Glue Grant and sponsored research agreements from Bristol Myers-Squibb, and Astrazeneca. TYW is supported by National Key R&D Program (Grant No: 2022YFC2502800), National Natural Science Fund of China (Grant No: 82388101) and Beijing Natural Science Foundation (Grant No: IS23096). DSWT is supported by the National Medical Research Council, Singapore (MOH-001689-00, MOH-000655-00 and MOH-001014-00), the Duke-NUS Medical School (Duke-NUS/RSF/2021/0018 and 05/FY2020/EX/15-A58, 05/FY2022/EX/66-A128) and the Agency for Science, Technology and Research (A20H4g2141 and H20C6a0032). D.S.J.T. acknowledges support from the Medical Research Council (MRC) Clinician Scientist Fellowship (UKRI2441), the MRC Proximity to Discovery and Impact (P2D) Fund (MR/X502996/1), and the Birmingham Health Partners Clinician Scientist Fellowship.
Conflicts of interest
DSWT holds a patent on a deep learning system for the detection of retinal diseases (10201706186 V) and a computer-implemented method for training an image classifier using weakly annotated training data (10201901083Y) and stocks at EyRIS, Singapore and aISIGHT. Dean P. Jones and Douglas I. Walker each hold the position of Editorial Board Member and did not participate in the peer-review or make any editorial decisions for this manuscript. The other authors declare no conflicts of interest.
Disclaimer
The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.
Data availability
No new data were generated or analyzed in support of this research.
References
1 ManolioTA, CollinsFS, CoxNJ, et al Finding the missing heritability of complex diseases. Nature 2009; 461:747–753. http://doi.org/10.1038/nature08494
2 MünzelT, SørensenM, HahadO, NieuwenhuijsenM, DaiberA. The contribution of the exposome to the burden of cardiovascular disease. Nat Rev Cardiol.Cardiol. 2023; 20:651–669. http://doi.org/10.1038/s41569-3–00873-3
3 Lefèvre-ArbogastS, ChakerJ, MercierF, et al Assessing the contribution of the chemical exposome to neurodegenerative disease. Nat Neurosci. 2024; 27:812–821. http://doi.org/10.1038/s41593-4–01627-1
4 MillerGW, JonesDP. The nature of nurture: refining the definition of the exposome. Toxicol Sci. 2014; 137:1–2. http://doi.org/10.1093/toxsci/kft251
5 VermeulenR, SchymanskiEL, BarabásiA-L, MillerGW. The exposome and health: where chemistry meets biology. Science. 2020; 367:392–396. http://doi.org/10.1126/science.aay3164
6 MillerGW, ConsortiumBE, BennettLM, et al, Banbury Exposomics Consortium Integrating exposomics into biomedicine. Science. 2025; 388:356–358. http://doi.org/10.1126/science.adr0544
7 ZhangY, GaoY, LiuQS, ZhouQ, JiangG. Chemical contaminants in blood and their implications in chronic diseases. J Hazard Mater. 2024; 466:133511. http://doi.org/10.1016/j.jhazmat.2024.133511
8 FäysF, HardyEM, PalazziP, HaanS, BeausoleilC, AppenzellerBMR. Biomonitoring of fast-elimination endocrine disruptors—results from a 6-month follow up on human volunteers with repeated urine and hair collection. Sci Total Environ. 2021; 778:146330. http://doi.org/10.1016/j.scitotenv.2021.146330
9 LaiY, KoelmelJP, WalkerDI, et al High-resolution mass spectrometry for human exposomics: expanding chemical space coverage. Environ Sci Technol. 2024; 58:12784–12822. http://doi.org/10.1021/acs.est.4c01156
10 BaccarelliA, DolinoyDC, WalkerCL. A precision environmental health approach to prevention of human disease. Nat Commun. 2023; 14:2449. http://doi.org/10.1038/s41467-3–37626-2
11 WuH, EckhardtCM, BaccarelliAA. Molecular mechanisms of environmental exposures and human disease. Nat Rev Genet. 2023; 24:332–344. http://doi.org/10.1038/s41576-2–00569-3
12 BellCG, LoweR, AdamsPD, et al DNA methylation aging clocks: challenges and recommendations. Genome Biol. 2019; 20:249. http://doi.org/10.1186/s13059-9–1824-y
13 ArgentieriMA, XiaoS, BennettD, et al Proteomic aging clock predicts mortality and risk of common age-related diseases in diverse populations. Nat Med. 2024; 30:2450–2460. http://doi.org/10.1038/s41591-4–03164-7
14 BEST (Biomarkers, EndpointS, and other Tools) Resource. 2016.
15 TingDSW, CheungCY-L, LimG, et al Development and validation of a deep learning system for diabetic retinopathy and related eye diseases using retinal images from multiethnic populations with diabetes. JAMA 2017; 318:2211–2223. http://doi.org/10.1001/jama.2017.18152
16 SevgiDD, SrivastavaSK, WhitneyJon, et al Characterization of ultra-widefield angiographic vascular features in diabetic retinopathy with automated severity classification. Ophthalmol. Sci 2021; 1:100049. http://doi.org/10.1016/j.xops.2021.100049
17 PrasannaP, BobbaV, FigueiredoN, et al Radiomics-based assessment of ultra-widefield leakage patterns and vessel network architecture in the PERMEATE study: insights into treatment durability. Br J Ophthalmol. 2021; 105:1155–1160. http://doi.org/10.1136/bjophthalmol-0–317182
18 MoosaviA, FigueiredoN, PrasannaP, et al Imaging features of vessels and leakage patterns predict extended interval aflibercept dosing using ultra-widefield angiography in retinal vascular disease: findings from the PERMEATE study. IEEE Trans Biomed Eng. 2021; 68:1777–1786. http://doi.org/10.1109/tbme.2020.3018464
19 ZhuZ, WangY, QiZ, et al Oculomics: current concepts and evidence. Prog Retin Eye Res. 2025; 106:101350. http://doi.org/10.1016/j.preteyeres.2025.101350
20 KeanePA, SaddaSR. Retinal imaging in the twenty-first century: state of the art and future directions. Ophthalmology 2014; 121:2489–2500. http://doi.org/10.1016/j.ophtha.2014.07.054
21 de CarloTE, RomanoA, WaheedNK, DukerJS. A review of optical coherence tomography angiography (OCTA). Int J Retina Vitreous 2015; 1:5. http://doi.org/10.1186/s40942-015-0005-8
22 ZhouY, WagnerSK, ChiaMA, et al AutoMorph: automated retinal vascular morphology quantification via a deep learning pipeline. Transl Vis Sci Technol 2022; 11:12–12. http://doi.org/10.1167/tvst.11.7.12
23 CheungCY, XuD, ChengC-Y, et al A deep-learning system for the assessment of cardiovascular disease risk via the measurement of retinal-vessel calibre. Nat Biomed Eng 2021; 5:498–508. http://doi.org/10.1038/s41551-0–00626-4
24 LiuX, ShenM, HuangS, LengL, ZhuD, LuF. Repeatability and reproducibility of eight macular intra-retinal layer thicknesses determined by an automated segmentation algorithm using two SD-OCT instruments. Plos ONE 2014; 9:e87996. http://doi.org/10.1371/journal.pone.0087996
25 GayraudL, MortamaisM, SchweitzerC, et al Association of long-term exposure to ambient air pollution with retinal neurodegeneration: the prospective Alienor study. Environ Res. 2023; 232:116364. http://doi.org/10.1016/j.envres.2023.116364
26 ProvostEB, Int PanisL, SaenenND, et al Recent versus chronic fine particulate air pollution exposure as determinant of the retinal microvasculature in school children. Environ Res. 2017; 159:103–110. http://doi.org/10.1016/j.envres.2017.07.027
27 Van AartCJC, NawrotTS, SioenI, et al Longitudinal association between psychosocial stress and retinal microvasculature in children and adolescents. Psychoneuroendocrinology 2018; 92:50–56. http://doi.org/10.1016/j.psyneuen.2018.03.022
28 ZekavatSM, RaghuVK, TrinderM, et al Deep learning of the retina enables phenome- and genome-wide analyses of the microvasculature. Circulation 2022; 145:134–150. http://doi.org/10.1161/CIRCULATIONAHA.121.057709
29 TodaN, Nakanishi-TodaM. How mental stress affects endothelial function. Pflugers Arch. 2011; 462:779–794. http://doi.org/10.1007/s00424-1–1022-6
30 HänselA, HongS, CámaraRJ, von KänelR. Inflammation as a psychophysiological biomarker in chronic psychosocial stress. Neurosci Biobehav Rev. 2010; 35:115–121. http://doi.org/10.1016/j.neubiorev.2009.12.012
31 GishtiO, JaddoeVW, FelixJF, et al Retinal microvasculature and cardiovascular health in childhood. Pediatrics 2015; 135:678–685. http://doi.org/10.1542/peds.4–3341
32 GopinathB, FloodVM, WangJJ, et al Carbohydrate nutrition is associated with changes in the retinal vascular structure and branching pattern in children. Am J Clin Nutr. 2012; 95:1215–1222. http://doi.org/10.3945/ajcn.111.031641
33 StreeseL, GueriniC, BühlmayerL, et al Physical activity and exercise improve retinal microvascular health as a biomarker of cardiovascular risk: a systematic review. Atherosclerosis 2020; 315:33–42. http://doi.org/10.1016/j.atherosclerosis.2020.09.017
34 ZekavatSM, JorsherySD, RauscherFG, et al Phenome- and genome-wide analyses of retinal optical coherence tomography images identify links between ocular and systemic health. Sci Transl Med. 2024; 16:eadg4517. http://doi.org/10.1126/scitranslmed.adg4517
35 WagnerSK, FuDJ, FaesL, et al Insights into systemic disease through retinal imaging-based oculomics. Transl Vis Sci Technol. 2020; 9:6–6. http://doi.org/10.1167/tvst.9.2.6
36 ZhangX, KongL, LiuS, et al EBD: an eye biomarker database. Bioinformatics 2023; 39. http://doi.org/10.1093/bioinformatics/btad194
37 ZhouY, ChiaMA, WagnerSK, et al; UK Biobank Eye & Vision Consortium A foundation model for generalizable disease detection from retinal images. Nature 2023; 622:156–163. http://doi.org/10.1038/s41586-3–06555-x
38 QiuJ, WuJ, WeiH, et al Development and validation of a multimodal multitask vision foundation model for generalist ophthalmic artificial intelligence. Nejm AI. 2024; 1:AIoa2300221. http://doi.org/10.1056/AIoa2300221
39 DhamdhereR, ModanwalG, MuthaP, et al; CRIC Study Investigators Fundus photograph-derived computational features predict risk of cardiovascular events in the chronic renal insufficiency cohort clinical observational study. Kidney360 2025;9900. http://doi.org/10.34067/kid.0000000955
40 SinghA, NookaAK, ModanwalG, et al AI-informed retinal biomarkers predict 10-year risk of onset of multiple hematological malignancies. Eur J Cancer.Cancer 2025; 229:115752. http://doi.org/10.1016/j.ejca.2025.115752
41 TanK-A, GuptaP, AgarwalA, et al State of science: choroidal thickness and systemic health. Surv Ophthalmol. 2016; 61:566–581. http://doi.org/10.1016/j.survophthal.2016.02.007
42 ChewEY, BurnsSA, AbrahamAG, et al Standardization and clinical applications of retinal imaging biomarkers for cardiovascular disease: a roadmap from an NHLBI workshop. Nat Rev Cardiol. 2025; 22:47–63. http://doi.org/10.1038/s41569-4–01060-8
43 VajaranantTS, HallakJ, EspelandMA, et al An association between large optic nerve cupping and cognitive function. Am J Ophthalmol. 2019; 206:40–47. http://doi.org/10.1016/j.ajo.2019.05.019
44 AshrafG, McGuinnessM, KhanMA, ObtinallaC, HadouxX, van WijngaardenP. Retinal imaging biomarkers of Alzheimer’s disease: a systematic review and meta-analysis of studies using brain amyloid beta status for case definition. Alzheimers Dement (Amst). 2023; 15:e12421. http://doi.org/10.1002/dad2.12421
45 WongCW, WongTY, ChengCY, SabanayagamC. Kidney and eye diseases: common risk factors, etiological mechanisms, and pathways. Kidney Int. 2014; 85:1290–1302. http://doi.org/10.1038/ki.2013.491
46 ZhuZ, ShiD, GuankaiP, et al Retinal age gap as a predictive biomarker for mortality risk. Br J Ophthalmol. 2023; 107:547–554. http://doi.org/10.1136/bjophthalmol-1–319807
47 ZhuZ, HuW, ChenR, et al Retinal age gap as a predictive biomarker of stroke risk. BMC Med. 2022; 20:466. http://doi.org/10.1186/s12916-2–02620-w
48 YuZ, ChenR, GuiP, et al A cross population study of retinal aging biomarkers with longitudinal pre-training and label distribution learning. Npj Digital Med 2025; 8:344. http://doi.org/10.1038/s41746-5–01751-7
49 LevineME, LuAT, QuachA, et al An epigenetic biomarker of aging for lifespan and healthspan. Aging (Albany NY) 2018; 10:573–591. http://doi.org/10.18632/aging.101414
50 NusinoviciS, RimTH, LiH, et al Application of a deep-learning marker for morbidity and mortality prediction derived from retinal photographs: a cohort development and validation study. Lancet Healthy Longev. 2024; 5:100593. http://doi.org/10.1016/S2666-7568(24)00089-8
51 SimMA, ThamYC, NusinoviciS, et al A deep-learning retinal aging biomarker for cognitive decline and incident dementia. Alzheimers Dement. 2025; 21:e14601. http://doi.org/10.1002/alz.14601
52 AminiP, OkemeJO. Tear fluid as a matrix for biomonitoring environmental and chemical exposures. Curr Environ Health Rep. 2024; 11:340–355. http://doi.org/10.1007/s40572-4–00454-7
53 WangJ, KangH, HuangX, LiuY, HeY, JieY. Identification of microplastics in human tear fluid and meibum: implications for dry eye disease pathogenesis. J Hazard Mater. 2025; 489:137635. http://doi.org/10.1016/j.jhazmat.2025.137635
54 ZhangK, YuL, QuL, et al Identifying and analyzing the microplastics in human aqueous humor by pyrolysis-gas chromatography/mass spectrometry. iScience. 2025; 28:112078. http://doi.org/10.1016/j.isci.2025.112078
55 ZhongY, YangY, ZhangL, et al Revealing new insights: two-center evidence of microplastics in human vitreous humor and their implications for ocular health. Sci Total Environ. 2024; 921:171109. http://doi.org/10.1016/j.scitotenv.2024.171109
56 JingD, JiangX, ZhouP, et al Evidence of air pollution-related ocular signs and altered inflammatory cytokine profile of the ocular surface in Beijing. Sci Rep. 2022; 12:18359. http://doi.org/10.1038/s41598-2–23294-7
57 WuD, LimBXH, SeahI, et al Impact of microplastics on the ocular surface. IJMS. 2023; 24:3928. http://doi.org/10.3390/ijms24043928
58 HongM, TongL, MehtaJS, OngHS. Impact of exposomes on ocular surface diseases. IJMS. 2023; 24:11273. http://doi.org/10.3390/ijms241411273
59 Fotovat-AhmadiN, SiddiquiO, OngJ, et al The ocular surface tear film as a biomarker for systemic health. Ocul Surf. 2025; 37:283–300. http://doi.org/10.1016/j.jtos.2025.05.005
60 BrunmairJ, BileckA, SchmidlD, et al Metabolic phenotyping of tear fluid as a prognostic tool for personalised medicine exemplified by T2DM patients. EPMA J. 2022; 13:107–123. http://doi.org/10.1007/s13167-2–00272-7
61 PhanN, LiY, YangM, LiuF. Tear fluid derived extracellular vesicles for new biomarker discovery. Ocul Surf. 2025; 37:314–322. http://doi.org/10.1016/j.jtos.2025.05.001
62 WolfJ, RasmussenDK, SunYJ, et al Liquid-biopsy proteomics combined with AI identifies cellular drivers of eye aging and disease. Cell 2023; 186:4868–4884.e12. http://doi.org/10.1016/j.cell.2023.09.012
63 ChungMK, HouseJS, AkhtariFS, et al; Members of the Exposomics Consortium Decoding the exposome: data science methodologies and implications in exposome-wide association studies (ExWASs). Exposome 2024; 4:osae001. http://doi.org/10.1093/exposome/osae001
64 LloydD, HouseJS, AkhtariFS, et al Questionnaire-based exposome-wide association studies for common diseases in the personalized environment and genes study. Exposome 2024; 4:osae002. http://doi.org/10.1093/exposome/osae002
65 ArgentieriMA, AminN, Nevado-HolgadoAJ, et al Integrating the environmental and genetic architectures of aging and mortality. Nat Med. 2025; 31:1016–1025. http://doi.org/10.1038/s41591-4–03483-9
66 LinBD, PriesLK, Arias-MagnascoA, et al Exposome-wide gene-by-environment interaction study of psychotic experiences in the UK Biobank. Biol Psychiatry Glob Open Sci. 2025; 5:100460. http://doi.org/10.1016/j.bpsgos.2025.100460
67 BobbJF, Claus HennB, ValeriL, CoullBA. Statistical software for analyzing the health effects of multiple concurrent exposures via Bayesian kernel machine regression. Environ Health. 2018; 17:67. http://doi.org/10.1186/s12940-8–0413-y
68 KeilAP, BuckleyJP, O'BrienKM, FergusonKK, ZhaoS, WhiteAJ. A quantile-based g-computation approach to addressing the effects of exposure mixtures. Environ Health Perspect. 2020; 128:47004. http://doi.org/10.1289/EHP5838
69 MigeotJ, Pina-EscuderoSD, HernandezH, et al Social exposome and brain health outcomes of dementia across Latin America. Nat Commun. 2025; 16:8196. http://doi.org/10.1038/s41467-5–63277-6
70 VanderWeeleTJ. Mediation analysis: a practitioner’s guide. Annu Rev Public Health. 2016; 37:17–32. http://doi.org/10.1146/annurev-publhealth-032315-021402
71 Chadeau-HyamM, AthersuchTJ, KeunHC, et al Meeting-in-the-middle using metabolic profiling—a strategy for the identification of intermediate biomarkers in cohort studies. Biomarkers 2011; 16:83–88. http://doi.org/10.3109/1354750x.2010.533285
72 MeiY, ChristensenGM, LiZ, et al Joint effects of air pollution and neighborhood socioeconomic status on cognitive decline—Mediation by depression, high cholesterol levels, and high blood pressure. Sci Total Environ. 2024; 923:171535. http://doi.org/10.1016/j.scitotenv.2024.171535
73 LiZ, ChristensenGM, LahJJ, et al Neighborhood characteristics as confounders and effect modifiers for the association between air pollution exposure and subjective cognitive functioning. Environ Res. 2022; 212:113221. http://doi.org/10.1016/j.envres.2022.113221
74 HuW, ChenS, CaiJ, YangY, YanH, ChenF. High-dimensional mediation analysis for continuous outcome with confounders using overlap weighting method in observational epigenetic study. BMC Med Res Methodol. 2024; 24:125. http://doi.org/10.1186/s12874-4–02254-x
75 PereraC, ZhangH, ZhengY, et al HIMA2: high-dimensional mediation analysis and its application in epigenome-wide DNA methylation data. BMC Bioinformatics. 2022; 23:296. http://doi.org/10.1186/s12859-2–04748-1
76 LiZ, LiangD, EbeltS, et al Differential DNA methylation in the brain as potential mediator of the association between traffic-related PM2.5 and neuropathology markers of Alzheimer’s disease. Alzheimers Dement. 2024; 20:2538–2551. http://doi.org/10.1002/alz.13650
77 LiZ, DunlopAL, SarnatJA, et al Unraveling the molecular links between fine particulate matter exposure and early birth risks in African American mothers: a metabolomics Study in the Atlanta African American Maternal-Child Cohort. Environ Sci Technol. 2025; 59:10905–10918. http://doi.org/10.1021/acs.est.5c02071
78 YazdiMD, WangY, DiQ, et al Long-term effect of exposure to lower concentrations of air pollution on mortality among US Medicare participants and vulnerable subgroups: a doubly-robust approach. Lancet Planet Health. 2021; 5:e689–e697. http://doi.org/10.1016/S2542-5196(21)00204-7
79 HébertJR, ShivappaN, WirthMD, HusseyJR, HurleyTG. Perspective: the dietary inflammatory index (DII)—lessons learned, improvements made, and future directions. Adv Nutr. 2019; 10:185–195. http://doi.org/10.1093/advances/nmy071
80 KarSS, AbrahamJ, WykoffCC, et al Computational imaging biomarker correlation with intraocular cytokine expression in diabetic macular edema: radiomics insights from the IMAGINE Study. Ophthalmol Sci. 2022; 2:100123. http://doi.org/10.1016/j.xops.2022.100123
81 AbrahamJR, WykoffCC, ArepalliS, et al Aqueous cytokine expression and higher order OCT biomarkers: assessment of the anatomic-biologic bridge in the IMAGINE DME Study. Am J Ophthalmol. 2021; 222:328–339. http://doi.org/10.1016/j.ajo.2020.08.047
82 ZhangK, ZhouR, AdhikarlaE, et al A generalist vision–language foundation model for diverse biomedical tasks. Nat Med. 2024; 30:3129–3141. http://doi.org/10.1038/s41591-4–03185-2
83 MoorM, BanerjeeO, AbadZSH, et al Foundation models for generalist medical artificial intelligence. Nature 2023; 616:259–265. http://doi.org/10.1038/s41586-3–05881-4
84 TeoZL, ThirunavukarasuAJ, ElangovanK, et al Generative artificial intelligence in medicine. Nat Med. 2025; 31:3270–3282. http://doi.org/10.1038/s41591-5–03983-2
85 LiY, HuangQ, JiangJ, et al Large language model-based biological age prediction in large-scale populations. Nat Med. 2025; 31:2977–2990. http://doi.org/10.1038/s41591-5–03856-8
86 ChenW, ZhangP, TranTN, et al A visual–omics foundation model to bridge histopathology with spatial transcriptomics. Nat Methods 2025; 22:1568–1582. http://doi.org/10.1038/s41592-5–02707-1
87 GazianoJM, ConcatoJ, BrophyM, et al Million Veteran Program: a mega-biobank to study genetic influences on health and disease. J Clin Epidemiol. 2016; 70:214–223. http://doi.org/10.1016/j.jclinepi.2015.09.016
88 TAoURPI. The “all of us” research program. N Engl J Med. 2019; 381:668–676. http://doi.org/10.1056/NEJMsr1809937
89 SudlowC, GallacherJ, AllenN, et al UK biobank: an open access resource for identifying the causes of a wide range of complex diseases of middle and old age. PLoS Med. 2015; 12:e1001779. http://doi.org/10.1371/journal.pmed.1001779
90 HubbardLD, BrothersRJ, KingWN, et al Methods for evaluation of retinal microvascular abnormalities associated with hypertension/sclerosis in the atherosclerosis risk in communities study. Ophthalmology 1999; 106:2269–2280. http://doi.org/10.1016/S0161-6420(99)90525-0
91 GrunwaldJE, AlexanderJ, MaguireM, et al; CRIC Study Group Prevalence of ocular fundus pathology in patients with chronic kidney disease. Clin J Am Soc Nephrol. 2010; 5:867–873. http://doi.org/10.2215/cjn.08271109
92 MaitreL, de BontJ, CasasM, et al Human Early Life Exposome (HELIX) study: a European population-based exposome cohort. BMJ Open. 2018; 8:e021311. http://doi.org/10.1136/bmjopen-7–021311
93 BrownSHJ, KunnenCME, PapasEB, et al Intersubject and interday variability in human tear and meibum lipidomes: a pilot study. Ocul Surf. 2016; 14:43–48. http://doi.org/10.1016/j.jtos.2015.08.005
94 JonesG, AltmanJ, AhmedS, et al Unraveling the intraday variations in the tear fluid proteome. Invest Ophthalmol Vis Sci. 2024; 65:2–2. http://doi.org/10.1167/iovs.65.3.2
95 MillerGW. Integrating exposomics into biomedicine. Science 2025; 388:356–358. http://doi.org/10.1126/science.adr054440273259
96 LloydD, HouseJS, AkhtariFS, et al Interactive data sharing for multiple questionnaire-based exposome-wide association studies and exposome correlations in the personalized environment and genes study. Exposome 2024; 4:osae003. http://doi.org/10.1093/exposome/osae003
97 LiuM, NingY, TeixayavongS, et al A translational perspective towards clinical AI fairness. Npj Digital Med. 2023; 6:172. http://doi.org/10.1038/s41746-3–00918-4
98 TeoZL, JinL, LiuN, et al Federated machine learning in healthcare: a systematic review on clinical applications and technical architecture. Cell Rep Med. 2024; 5:101481. http://doi.org/10.1016/j.xcrm.2024.101419
99 SchmittCP, StingoneJA, RajasekarA, et al; A roadmap to advance exposomics through federation of data. Exposome. 2023; 3. http://doi.org/10.1093/exposome/osad010
100 ThamYC, GohJHL, NderituP, et al; Global RETFound Consortium Building the world’s first truly global medical foundation model. Nat Med. 2025. http://doi.org/10.1038/s41591-5–03859-5
101 ChuaSYL, KhawajaAP, DickAD, et al; UK Biobank Eye and Vision Consortium Ambient air pollution associations with retinal morphology in the UK Biobank. Invest Ophthalmol Vis Sci. 2020; 61:32. http://doi.org/10.1167/iovs.61.5.32
102 KorsiakJ, PerepelukK-L, PetersonNG, KulkaR, WeichenthalS. Air pollution and retinal vessel diameter and blood pressure in school-aged children in a region impacted by residential biomass burning. Sci Rep. 2021; 11:12790. http://doi.org/10.1038/s41598-1–92269-x
103 AdarSD, KleinR, KleinB, et al Long-term exposures to air pollution and retinal micro-vascular caliber: the multi-ethnic study of atherosclerosis (MESA). Epidemiology 2008; 19:S337–S338. http://doi.org/10.1097/01.ede.0000340544.93470.e3
104 Özata GündoğduK, DoğanE, ÇelikE, AlagözG. Retinal nerve fiber layer and ganglion cell complex thickness in diabetic smokers without diabetic retinopathy. Cutan Ocul Toxicol. 2024; 43:22–26. http://doi.org/10.1080/15569527.2023.2268162
105 CheungCY, ZhangXJ, ChanH-N, et al Influence of secondhand smoke exposure on the retinal vasculature of children in Hong Kong. Commun Med (Lond) 2023; 3:155. http://doi.org/10.1038/s43856-3–00389-4
106 LiewG, SharrettAR, WangJJ, et al Relative importance of systemic determinants of retinal arteriolar and venular caliber: the atherosclerosis risk in communities study. Arch Ophthalmol. 2008; 126:1404–1410. http://doi.org/10.1001/archopht.126.10.1404
107 ZhuX, YangK, XiaoY, et al Association of cigarette smoking with retinal capillary plexus: an optical coherence tomography angiography study. Acta Ophthalmol. 2022; 100:e1479–e1488. http://doi.org/10.1111/aos.15157
108 Quiroz-ReyesMA, Quiroz-GonzalezEA, Quiroz-GonzalezMA, Lima-GomezV. Effects of cigarette smoking on retinal thickness and choroidal vascularity index: a systematic review and meta-analysis. Int J Retina Vitreous. 2025; 11:21. http://doi.org/10.1186/s40942-5–00646-9
109 KanH, StevensJ, HeissG, KleinR, RoseKM, LondonSJ. Dietary fiber intake and retinal vascular caliber in the Atherosclerosis Risk in Communities Study2. Am J Clin Nutr. 2007; 86:1626–1632. http://doi.org/10.1093/ajcn/86.5.1626
110 RepoO, JuonalaM, NiinikoskiH, et al Randomized 20-year infancy-onset dietary intervention, life-long cardiovascular risk factors and retinal microvasculature. Eur Heart J. 2024; 45:3072–3085. http://doi.org/10.1093/eurheartj/ehae423
111 VasileiouES, FilippatouAG, Pimentel MaldonadoD, et al Socioeconomic disparity is associated with faster retinal neurodegeneration in multiple sclerosis. Brain 2021; 144:3664–3673. http://doi.org/10.1093/brain/awab342
112 KöchliS, EndesK, GrenacherJ, et al Socioeconomic status and parental lifestyle are associated with vascular phenotype in children. Front Public Health. 2021; 9:610268. http://doi.org/10.3389/fpubh.2021.610268
113 CuiB, ZhuY, ZhangX, et al Association of physical activity with retinal thickness and vascular structure in elderly chinese population. Ophthalmic Res. 2023; 66:281–292. http://doi.org/10.1159/000527448
114 BerryEC, MarshallHN, MullanyS, et al Physical activity is associated with macular thickness: a multi-cohort observational study. Invest Ophthalmol Vis Sci. 2023; 64:11. http://doi.org/10.1167/iovs.64.3.11
115 ZhangXJ, YuenVL, ZhangY, et al Effects of physical activity and inactivity on microvasculature in children: the hong kong children eye study. Invest Ophthalmol Vis Sci. 2024; 65:7. http://doi.org/10.1167/iovs.65.14.7
116 HaoR, ZhangM, ZhaoL, et al Impact of air pollution on the ocular surface and tear cytokine levels: a multicenter prospective cohort study. Front Med (Lausanne) 2022; 9:909330. http://doi.org/10.3389/fmed.2022.909330
117 MatsudaM, BonattiR, MarqueziniMV, et al Lacrimal cytokines assessment in subjects exposed to different levels of ambient air pollution in a large metropolitan area. Plos ONE 2015; 10:e0143131. http://doi.org/10.1371/journal.pone.0143131
118 GargA, RewriP. Comparison of aqueous humor ascorbic acid level in smokers and non-smokers. Exp Eye Res. 2023; 226:109302. http://doi.org/10.1016/j.exer.2022.109302
119 AmerR, KoriatA. Aqueous humor perturbations in chronic smokers: a proteomic study. Sci Rep. 2024; 14:11279. http://doi.org/10.1038/s41598-4–62039-6
120 AlkoziHA, AlhudhayfHA, AlawadNMA. Association between dry eye disease with anxiety and depression among medical sciences students in qassim region: cortisol levels in tears as a stress biomarker. J Multidiscip Healthc. 2024; 17:4549–4557. http://doi.org/10.2147/jmdh.S488956
121 SunC, ChenX, HuangY, et al Effects of aerobic exercise on tear secretion and tear film stability in dry eye patients. BMC Ophthalmol. 2022; 22:9. http://doi.org/10.1186/s12886-1–02230-9
122 PapathanasiouKA, KazantzisD, VrachatisDA, et al Choroidal thickness in patients with systemic arterial hypertension: a systematic review and meta-analysis. Ther Adv Ophthalmol. 2022; 14:25158414221132825. http://doi.org/10.1177/25158414221132825
123 TanW, YaoX, LeTT, et al The application of optical coherence tomography angiography in systemic hypertension: a meta-analysis. Front Med (Lausanne). 2021; 8:778330. http://doi.org/10.3389/fmed.2021.778330
124 CheungCY, ZhengY, HsuW, et al Retinal vascular tortuosity, blood pressure, and cardiovascular risk factors. Ophthalmology 2011; 118:812–818. http://doi.org/10.1016/j.ophtha.2010.08.045
125 WongTY, IslamFMA, KleinR, et al Retinal vascular caliber, cardiovascular risk factors, and inflammation: the multi-ethnic study of atherosclerosis (MESA). Invest Ophthalmol Vis Sci. 2006; 47:2341–2350. http://doi.org/10.1167/iovs.5–1539
126 SmithW, WangJJ, WongTY, et al Retinal Arteriolar narrowing is associated with 5-year incident severe. Hypertension. 2004; 44:442–447. http://doi.org/10.1161/01.HYP.0000140772.40322.ec
127 XiaoW, GuoX, DingX, HeM. Serum lipid profiles and dyslipidaemia are associated with retinal microvascular changes in children and adolescents. Sci Rep. 2017; 7:44874. http://doi.org/10.1038/srep44874
128 CourtieE, KirkpatrickJRM, TaylorM, et al Optical coherence tomography angiography analysis methods: a systematic review and meta-analysis. Sci Rep. 2024; 14:9643. http://doi.org/10.1038/s41598-4–54306-3
129 van DijkHW, VerbraakFD, KokPHB, et al Decreased retinal ganglion cell layer thickness in patients with type 1 diabetes. Invest Ophthalmol Vis Sci. 2010; 51:3660–3665. http://doi.org/10.1167/iovs.9–5041
130 SeidelmannSB, ClaggettB, BravoPE, et al Retinal vessel calibers in predicting long-term cardiovascular outcomes. Circulation 2016; 134:1328–1338. http://doi.org/10.1161/CIRCULATIONAHA.116.023425
131 LyeWK, PatersonE, PattersonCC, et al A systematic review and participant-level meta-analysis found little association of retinal microvascular caliber with reduced kidney function. Kidney Int. 2021; 99:696–706. http://doi.org/10.1016/j.kint.2020.06.033
132 WangL, WangJ-Y, ChenC, et al OCTA evaluates changes in retinal microvasculature in renal hypertension patients. Sci Rep. 2024; 14:28910. http://doi.org/10.1038/s41598-4–68690-3
133 LemmensS, DevulderA, Van KeerK, BierkensJ, De BoeverP, StalmansI. Systematic review on fractal dimension of the retinal vasculature in neurodegeneration and stroke: assessment of a potential biomarker. systematic review. Front Neurosci. 2020; 14:16. http://doi.org/10.3389/fnins.2020.00016
134 CostanzoE, LengyelI, ParravanoM, et al Ocular biomarkers for Alzheimer disease dementia: an umbrella review of systematic reviews and meta-analyses. JAMA Ophthalmol. 2023; 141:84–91. http://doi.org/10.1001/jamaophthalmol.2022.4845
135 ChanVTT, SunZ, TangS, et al Spectral-domain OCT measurements in Alzheimer’s disease: a systematic review and meta-analysis. Ophthalmology 2019; 126:497–510. http://doi.org/10.1016/j.ophtha.2018.08.009
136 CheungCY, RanAR, WangS, et al A deep learning model for detection of Alzheimer’s disease based on retinal photographs: a retrospective, multicentre case-control study. Lancet Digit Health 2022; 4:e806–e815. http://doi.org/10.1016/s2589-7500(22)00169-8
137 IkramMA, BrusselleGGO, MuradSD, et al The Rotterdam Study: 2018 update on objectives, design and main results. Eur J Epidemiol. 2017; 32:807–850. http://doi.org/10.1007/s10654-7–0321-4
138 ChenZ, ChenJ, CollinsR, et al; China Kadoorie Biobank (CKB) collaborative group China Kadoorie Biobank of 0.5 million people: survey methods, baseline characteristics and long-term follow-up. Int J Epidemiol. 2011; 40:1652–1666. http://doi.org/10.1093/ije/dyr120
139 ReicherL, ShiloS, GodnevaA, et al Deep phenotyping of health–disease continuum in the Human Phenotype Project. Nat Med. 2025; 31:3191. http://doi.org/10.1038/s41591-5–03790-9
140 WongE, BertinN, HebrardM, et al; SG10K_Health Consortium The Singapore national precision medicine strategy. Nat Genet. 2023; 55:178–186. http://doi.org/10.1038/s41588-2–01274-x
141 OwsleyC, MatthiesDS, McGwinG, et al; AI-READI Consortium Cross-sectional design and protocol for Artificial Intelligence Ready and Equitable Atlas for Diabetes Insights (AI-READI). BMJ Open. 2025; 15:e097449. http://doi.org/10.1136/bmjopen-4–097449
142 INVESTIGATORS TA. The atherosclerosis risk in communit (ARIC) study: design and objectives. Am J Epidemiol. 1989; 129:687–702. http://doi.org/10.1093/oxfordjournals.aje.a115184
143 MajithiaS, ThamY-C, CheeM-L, et al Cohort profile: the Singapore epidemiology of eye diseases study (SEED). Int J Epidemiol. 2021; 50:41–52. http://doi.org/10.1093/ije/dyaa238
144 KleinR, KleinBE, LintonKL, De MetsDL. The Beaver Dam eye study: visual acuity. Ophthalmology. 1991; 98:1310–1315. http://doi.org/10.1016/s0161-6420(91)32137-7
145 AtteboK, MitchellP, SmithW. Visual acuity and the causes of visual loss in Australia: the blue mountains eye study. Ophthalmology 1996; 103:357–364. http://doi.org/10.1016/S0161-6420(96)30684-2
146 BildDE, BluemkeDA, BurkeGL, et al Multi-ethnic study of atherosclerosis: objectives and design. Am J Epidemiol. 2002; 156:871–881. http://doi.org/10.1093/aje/kwf113
147 JonasJB, XuL, WangYX. The Beijing eye study. Acta Ophthalmol. 2009; 87:247–261. http://doi.org/10.1111/j.5–3768.2008.01385.x
148 FeldmanHI, AppelLJ, ChertowGM, et al; Chronic Renal Insufficiency Cohort (CRIC) Study Investigators The chronic renal insufficiency cohort (CRIC) study: design and methods. J Am Soc Nephrol. 2003; 14:S148–53. http://doi.org/10.1097/01.asn.0000070149.78399.ce