Emerging applications of bioinformatics and artificial intelligence in the analysis of biofluid markers involved in retinal occlusive diseases: a systematic review.

PURPOSE

To review the literature on the application of bioinformatics and artificial intelligence (AI) for analysis of biofluid biomarkers in retinal vein occlusion (RVO) and their potential utility in clinical decision-making.

METHODS

We systematically searched MEDLINE, Embase, Cochrane, and Web of Science databases for articles reporting on AI or bioinformatics in RVO involving biofluids from inception to August 2021. Simple AI was categorized as logistics regressions of any type. Risk of bias was assessed using the Joanna Briggs Institute Critical Appraisal Tools.

RESULTS

Among 10,264 studies screened, 14 eligible articles, encompassing 578 RVO patients, met the inclusion criteria. The use and reporting of AI and bioinformatics was heterogenous. Four articles performed proteomic analyses, two of which integrated AI tools such as discriminant analysis, probabilistic clustering, and string pathway analysis. A metabolomic study used AI tools for clustering, classification, and predictive modeling such as orthogonal partial least squares discriminant analysis. However, most studies used simple AI (n = 9). Vitreous humor sample levels of interleukin-6 (IL-6), vascular endothelial growth factor (VEGF), and aqueous humor levels of intercellular adhesion molecule-1 and IL-8 were implicated in the pathogenesis of branch RVO with macular edema. IL-6 and VEGF may predict visual acuity after intravitreal injections or vitrectomy, respectively. Metabolomics and Kyoto Encyclopedia of Genes and Genomes enrichment analysis identified the metabolic signature of central RVO to be related to lower aqueous humor concentration of carbohydrates and amino acids. Risk of bias was low or moderate for included studies.

CONCLUSION

Bioinformatics has applications for analysis of proteomics and metabolomics present in biofluids in RVO with AI for clinical decision-making and advancing the future of RVO precision medicine. However, multiple limitations such as simple AI use, small sample volume, inconsistent feasibility of office-based sampling, lack of longitudinal follow-up, lack of sampling before and after RVO, and lack of healthy controls must be addressed in future studies.