Researchers recently utilized artificial intelligence to analyze high-resolution retinal thickness data, which revealed novel genetic loci and systemic disease associations that may impact retinal health.
The study leveraged optical coherence tomography imaging data from the UK Biobank to process over 85,000 retinal scans and analyze fine-scale variations in macular thickness.
Led by V. E. Jackson, of the Walter and Eliza Hall Institute of Medical Research and the University of Melbourne in Australia, the researchers applied a deep convolutional neural network to segment retinal thickness (RT) across 29,000 macular points in 54,844 participants. They then analyzed genetic associations, metabolic influences, and systemic disease correlations with retinal thickness.
The parafoveal region was found to be particularly sensitive to systemic diseases, supporting the hypothesis that retinal imaging may serve as a biomarker for broader neurologic and vascular conditions.
According to their work, published in Nature Communications, nearly 300 genetic loci were associated with retinal thickness. Four loci on the X chromosome were identified, including SHROOM2 (rs554433) and EFNB1 (rs626840)—genes that have previously been linked to eye pigmentation and optic nerve abnormalities. The researchers also confirmed 64% of previously identified genetic associations and identified 123 novel loci.
“Further understanding of our genetic results will require single-cell and spatial transcriptomics of the retina to investigate specific cell types being affected as well as determine biological mechanisms … [and] will extend our spatial approach to the retinal sublayers, allowing us to further tease apart many of the associations and explore cell type specific effects,” the study authors wrote.
The researchers linked retinal thinning to systemic diseases such as multiple sclerosis (MS), type 2 diabetes (T2D), hypertension, and hyperlipidemia. Patients with MS exhibited significant retinal thinning, particularly in the nasal perifoveal region, which correlated with optic nerve involvement. The researchers suggested that “future studies utilizing retinal thickness as a biomarker should focus on [this area … because [it] contains the greatest signal.” Hypertension and T2D were among the strongest predictors of retinal thinning and they supported evidence that metabolic dysfunction may be linked to retinal health.
Specifically, lower low-density lipoprotein (LDL) cholesterol, very LDL, and apoB levels were associated with retinal thinning. Higher glucose, triglycerides, and branched-chain amino acids were linked to reduced retinal thickness as well, particularly in the parafoveal region. Omega-3 fatty acids were negatively associated with retinal thickness, a finding that the researchers noted requires further investigation.
In blood and inflammatory biomarkers, higher leukocyte and neutrophil counts were associated with parafoveal retinal thinning, and reticulocyte abnormalities correlated with retinal thickness changes.
Limitations included the need to impute missing data points and to exclude loci with specific criteria as well as underpowered analyses of participants with non-European ancestry.
“Our findings aim to direct future biomarker studies and biological experiments. This work validated and refined previous associations and will encourage others to reprocess high-dimensional data with [artificial intelligence],” the study authors concluded.
A full list of disclosures can be found in the published study.