Abstract
Purpose:
To differentiate between healthy and disease states by using a novel approach in which a subject’s cytokine composition is represented as a vector in multi-dimensional space. Using this method, novel biomarker equations are developed that could distinguish control subjects from those with proliferative diabetic retinopathy (PDR) and neovascular glaucoma (NVG). <br />
Methods:
Vitreous samples were collected from control subjects (n=28), PDR (n=35), and NVG secondary to PDR (n=4) during routine vitrectomy procedure. Control vitreous included those from epiretinal membranes, macular holes, and vitreomacular traction. Multiplex analysis was performed for detection of 31 pro-angiogenic and pro-inflammatory cytokines. Data was subjected to novel statistical methods in which the cumulative composition of each subject’s cytokine levels were represented as a vector in multi-dimensional space. Biomarker equations were constructed based on this analysis.
Results:
In subjects with PDR, cumulative vectors had distinct characteristics in multi-dimensional space with chi-squared values significantly greater than controls in 97% of cases (P<0.01). In NVG, multi-dimensional vectors showed chi-squared values that were even further separated from control and PDR (P<0.01). Using key cytokines that significantly contribute to separation of these vectors, a biomarker equation was constructed that can predict PDR in any individual with high certainty.
Conclusions:
We report a novel method by representing a collective biomarker composition of a subject as a single vector in a multi-dimensional space. This methodology can distinguish subjects with PDR and NVG from control with high certainty. Using this model, biomarker equations could be developed to distinguish healthy from disease states with high predictability. <br />