A simple blood test could soon stop diabetes from stealing people's sight before they even notice the damage. Researchers in Guangzhou, China have developed an AI model that reads 71 different proteins in the blood to predict diabetic retinal neurodegeneration—a condition that causes vision loss in diabetic patients—years before symptoms appear, according to a study published in PLOS Medicine by Huangdong Li and colleagues at the Guangdong Provincial Clinical Research Center for Ocular Diseases.
The scale of the problem is staggering. More than half a billion people worldwide now live with diabetes, and many face a silent threat to their vision. Diabetic retinal neurodegeneration damages the retina, the light-detecting tissue at the back of the eye, causing severe visual impairment and blindness. What makes this particularly urgent is that scientists now view DRN as a window into broader diabetic damage—the same degenerative processes can harm other parts of the nervous system, contributing to cognitive decline, dementia, and nerve damage in the extremities. Yet today, doctors can only diagnose DRN after the damage becomes visible on scans, meaning intervention comes far too late.
Li's team took a different approach. They analyzed blood plasma from 1,492 patients with type 2 diabetes in the Guangzhou Diabetic Eye Study who did not yet have DRN, tracking the eyesight of 1,218 of them over six years through detailed retinal scans. They validated their findings against 502 additional diabetic patients in the United Kingdom BioBank. Through this meticulous work, they identified 71 blood proteins linked to DRN risk—proteins involved in inflammation and cellular maintenance pathways that reveal early warning signs long before the eye shows damage.
Using machine learning, the researchers built a predictive model called Pro-DRN that outperformed the best existing model by 26 percent. The breakthrough lies in its simplicity and accessibility: a blood sample analyzed by AI can now flag which diabetic patients are at highest risk of retinal degeneration, allowing doctors to shift from reactive treatment to proactive prevention. The team has already made Pro-DRN available online so physicians can use it immediately in clinical practice.
The implications are profound. By moving diabetic eye care away from detecting established damage and toward earlier, molecular-level risk assessment, doctors can target closer monitoring and future neuroprotective treatments to the patients most likely to benefit. Instead of waiting for vision loss to announce itself, clinicians can now see the molecular shadows cast by emerging disease and act while the retina can still be saved.
The authors note that while Pro-DRN reveals associations between blood proteins and retinal damage rather than proving direct causation, it opens a new window into prevention. Early retinal nerve damage in diabetes leaves measurable signals in the blood—signals that, until now, have gone unread. For the hundreds of millions of people managing diabetes worldwide, this small vial of blood could mean the difference between a lifetime of sight and a lifetime of darkness.