Association between diabetic retinopathy and interleukin-related gene polymorphisms: a machine learning aided meta-analysis.

Background: Diabetic retinopathy (DR) is a severe complication of diabetes and a common cause of visual loss in adults. We aimed to assess the correlation between IL gene-related SNPs and the incidence of DR and attempted to predict DR with combined mutation site detection.

Methods: A systematic search of databases was performed up to August 2019. Five genetic models were used to analyze associations. Machine learning methods were implemented to improve SNP-related disease prediction.

Results: Sixteen trials assessing a total of 7221 patients were included in our meta-analysis. IL6/rs1800795, rs1800796, and IL10/rs1800896 were analyzed. For the IL-6 gene, there was no significant association between rs1800795 and the incidence of DR (allelic model: OR, 1.091; 95% CI, 0.892-1.334; = .396). There was no significant correlation between rs1800796 (allelic model: OR, 1.135; 95% CI, 0.678-1.901; = .63), rs1800896 (allelic model: OR, 1.047; 95% CI, 0.788-1.392; = .752) and the incidence of DR. Unfortunately, the machine learning results also showed that the combined detection of two SNPs could not accurately predict DR occurrence.

Conclusion: rs1800795 and rs1800796 in the IL-6 gene and rs1800896 in IL-10 gene are not related to the incidence of DR. Mutations in multiple SNPs for each DR patient still need to be specifically assessed to increase prediction accuracy.