Abstract
INTRODUCTION: Diabetic retinopathy (DR) is the most common chronic complication of diabetes, the leading cause of vision impairments in working-aged adults, and a significant cause of reduced quality of life for diabetic patients. Diabetic patients are recommended to have regular screening in order to catch DR at an early enough stage for effective management. However, due to a variety of factors, many patients can still fall through the cracks with the current screening methods. METHODS: Several long non-coding RNAs (lncRNAs), essential regulators of physiological and pathological processes, were previously identified by us as potential markers for DR phenotypes. In this study, we used a significantly larger sample set to validate our panel of lncRNAs. We also explored the possibility of creating a statistical model to detect DR from serum samples using the expression profiles of these lncRNAs. RESULTS: Our regression models, based solely on lncRNA expression data, demonstrated the ability to adequately detect DR and potentially predict it. Models based solely on lncRNA expression performed equally or better compared to models with additional patient information. The models showed promising performance, suggesting that serum lncRNA expression profiles could serve as reliable markers for DR detection. DISCUSSION: Further longitudinal studies are necessary to validate the model's capability to predict retinopathy in diabetic patients not yet diagnosed with DR. Nevertheless, our findings indicate that this lncRNA panel may offer a viable option for a simple, accessible, and convenient blood-based screening test for DR.