Development of targeted drugs for diabetic retinopathy using Mendelian randomized pharmacogenomics.

Purpose: This study aims to utilize genetic instrumental variables - protein quantitative trait loci (pQTL), and through analysis methods such as Mendelian randomization (MR), systematically screen and validate druggable proteins that have a causal relationship with diabetic retinopathy (DR), and further explore related drug targets, providing genetic evidence and new directions for the drug development of this disease. Methods: The research was based on large-scale public databases to conduct two-sample Mendelian randomization (MR) analysis. Firstly, 511 encoded proteins were selected from the known 4,479 druggable genes as initial exposure factors, with the summary data of GWAS for diabetic retinopathy as the outcome. MR analysis was conducted using the inverse variance weighted (IVW) method and the Wald ratio method, and strict screening was performed through Bonferroni correction. For the significantly associated proteins, heterogeneity tests, pleiotropy tests, leave-one-out analysis, and Steiger directionality tests were further conducted to verify the robustness of the results. Additionally, summary MR (SMR) analysis and colocalization analysis (coloc) were used to confirm the reliability of the causal relationship. Finally, a protein-protein interaction (PPI) network was constructed using the STRING database, and potential targeted drugs were mined from the DrugBank and DSigDB databases. Results: A preliminary analysis identified 37 proteins with potential causal relationships to DR (p < 0.05). After more rigorous pQTL screening and multiple testing corrections, it was found that Noggin (NOG) protein has a significant negative causal relationship with the risk of DR (p.adjust < 0.05), meaning that higher NOG protein levels may reduce the risk of disease. All sensitivity analyses supported the robustness of this result (no heterogeneity, no pleiotropy), and SMR and colocalization analyses (PP.H4 > 0.8) further confirmed this causal association. PPI network analysis revealed that NOG interacts with 10 proteins (such as BMP2, BMP4, etc.). Drug mining identified DB01373 as a corresponding drug for BMP4, and through DSigDB analysis, progesterone and estradiol were found to be potential therapeutic compounds targeting the NOG network. Conclusions: Through comprehensive genetic analysis, this study identified the NOG protein as a novel potential protective drug target for DR. Its function may be achieved by regulating the BMP signaling pathway. The research findings not only provide a new perspective for understanding the pathogenesis of this disease but also recommend existing drugs such as progesterone and estradiol as potential therapies, which are worthy of further functional experiments and clinical studies for verification.