Background: Diabetic retinopathy (DR) is a leading cause of blindness, with platelet dysfunction increasingly implicated. However, establishing a causal role and underlying mechanisms remains challenging. Methods: We employed an integrated multi-omics approach. Bidirectional two-sample Mendelian randomization (MR) using large-scale GWAS data (FinnGen R9, n=351,387) assessed causality between platelet indices and DR. Transcriptomic datasets (GSE94019, GSE102485) identified platelet-related genes (PRGs) via WGCNA and differential expression. Diagnostic signatures were refined using machine learning (Boruta, RF, LASSO). Single-cell RNA sequencing (GSE248284) revealed platelet-immune crosstalk. MR analysis also investigated causal effects of PRG expression (eQTLGen, n=31,684) on platelet indices. Computational pharmacology leveraging traditional Chinese medicine databases identified potential therapeutic compounds. Results: Forward MR identified platelet distribution width (PDW, OR=40.59, p<0.001) and mean platelet volume (MPV, OR=40.86, p<0.001) as significant causal risk factors for DR. Reverse MR showed higher expression of FTL (OR=0.96) and ATP2B2 (OR=0.96) attenuated platelet activation, while THBD increased risk (OR=1.01). A three-gene diagnostic panel (ATP2B2, FTL, THBD) achieved perfect accuracy (AUC=1.0). Single-cell analysis revealed distinct platelet subpopulations and dynamic crosstalk with immune cells via TGF-β, PDGF, and CCL signaling. Molecular docking identified folic acid as a high-affinity modulator of ATP2B2 (binding energy: -8.3 kcal/mol). Conclusion: This study establishes a bidirectional causal link between platelet indices and DR, identifying FTL, ATP2B2, and THBD as key mediators. The diagnostic panel and folate therapy offer novel strategies, repositioning platelets as active drivers of DR pathogenesis.