Mendelian Randomization of the Proteome: Genetic Evidence for Drug Targets in Vision-Threatening Diabetic Retinopathy

Diabetic retinopathy affects approximately 30–40% of people with diabetes globally. Making diabetic retinopathy one of the most common and serious complications of diabetes that threatens vision and significantly reduces quality of life for many patients. As the disease progresses, patients often experience worsening visual impairment, which impacts daily activities, independence, and mental well-being. In 2020 alone, diabetic retinopathy caused 1.07 million cases of blindness and nearly 3.28 million cases of moderate to severe visual impairment worldwide, contributing to a substantial and rising global burden, particularly among working-age adults in low- and middle-income countries [1, 2].This study represents a noteworthy effort to leverage genetic epidemiology for identifying novel therapeutic targets in vision-threatening diabetic retinopathy (VTDR). The study uses Mendelian randomization (MR) with plasma protein quantitative trait loci (pQTLs) as instruments to test whether circulating proteins causally influence VTDR risk. This approach reduces bias from confounding and reverse causation that typically affect observational studies.The integration of large-scale genome-wide association study (GWAS) data from diverse European populations enhances the robustness of findings, with rigorous quality controls (such as excluding datasets with missing data or poor SNP coverage) ensuring data integrity. Notably, the analysis restricts SNPs to those with p<5×10⁻⁸ for strong instrument relevance, and standardization of alleles clarifies causal directions. The proteome-wide, hypothesis-free approach screens hundreds of proteins, potentially uncovering unconventional targets beyond established pathways like VEGF. Innovative elements include bidirectional MR for directionality and sensitivity analyses for pleiotropy to contextualize hits like MMP8 and BST1 in proliferative diabetic retinopathy (PDR) pathogenesis. Good visualization of analytical steps further aids transparency, while aggregating multiple datasets and risk factors offers deeper mechanistic insights into VTDR.Despite these strengths, some limitations warrant consideration. The term "vision-threatening diabetic retinopathy" lacks standardization as a MeSH term, potentially limiting the comprehensiveness of the literature review on risk factors; expanding searches with synonyms like "severe diabetic retinopathy" or "proliferative retinopathy" could yield more relevant studies. The authors acknowledge MR assumptions such as independence from confounders and absence of horizontal pleiotropy. However, these conditions may not fully hold because SNPs can tag nearby but unrelated causal variants, and proteins may influence VTDR through unmeasured intermediates like lipids or blood pressure. This warrants multivariable MR and rigorous colocalization analyses. Lifelong genetic effects captured by MR may not mirror drug interventions in adults, where compensatory mechanisms differ. The PPI network's role in screening MMP8 and BST1 remains unclear, lacking clear methodology on how interactions informed target selection. Although DrugBank and DGIdb were queried without identifying drugs, using additional databases might reinforce confidence in target druggability.In future studies, replication in independent cohorts and wet-lab validation will be crucial, especially given the tissue mismatch between plasma and ocular environments such as the retina and vitreous. Ancestry-limited European datasets may underestimate effects in diverse populations where VTDR prevalence varies. Integrating multi-omics data and time-resolved analyses could sharpen mechanistic insight and pave the way for personalized therapies in this sight-threatening complication of diabetes. Overall, the study advances drug-repurposing prospects but underscores the need for cautious interpretation and further empirical confirmation. Finally, the biological consequences of altered proteins on relevant pathways warrant deeper exploration beyond association signals.