Various immune mediators have crucial roles in the pathogenesis of intraocular diseases. Machine learning can be used to automatically select and weigh various predictors to develop models maximizing predictive power. However, these techniques have not yet been extensively applied studies focused on intraocular diseases. We evaluated whether five machine learning algorithms applied to the data of immune mediator levels in aqueous humor can accurately predict the actual diagnoses of 17 selected intraocular diseases, and identified which immune mediators drive the predictive power of a machine learning model. Cross-sectional study. 512 eyes with diagnoses of 17 intraocular diseases. Aqueous humor samples were collected, and the concentrations of 28 immune mediators were determined using a cytometric bead array. Each immune mediator was ranked according to its importance using five machine learning algorithms: random forest (RF), linear support vector machine (SVM), radial basis function SVM, decision tree and naïve Bayes classifier. Stratified k-fold cross-validation was used in evaluation of algorithms with dataset divided into training and test datasets. The algorithms were evaluated in terms of precision, recall, accuracy, F-score, area under the receiver operating characteristics curve, area under the precision-recall curve and mean decrease in Gini index. Among the five machine learning models, RF yielded the highest classification accuracy in multi-class differentiation of 17 intraocular diseases. The RF prediction models for vitreoretinal lymphoma, acute retinal necrosis, endophthalmitis, rhegmatogenous retinal detachment, and primary open angle glaucoma achieved the highest classification accuracy, precision, and recall. RF recognized vitreoretinal lymphoma, acute retinal necrosis, endophthalmitis, rhegmatogenous retinal detachment and primary open angle glaucoma with the top five F-scores. The three highest-ranking relevant immune mediators were IL-10, IP-10 and angiogenin for prediction of vitreoretinal lymphoma; Mig, IFN-γ and IP-10 for acute retinal necrosis; and IL-6, G-CSF and IL-8 for endophthalmitis. RF algorithms based on 28 immune mediators in aqueous humor successfully predicted the diagnosis of vitreoretinal lymphoma, acute retinal necrosis, and endophthalmitis. Overall, the findings of the present study contribute to increased knowledge on new biomarkers that can potentially facilitate diagnosis of intraocular diseases in the future.

The activities of various metabolic pathways can influence the pathogeneses of autoimmune diseases, and intrinsic metabolites can potentially be used to diagnose diseases. However, the metabolomic analysis of patients with uveitis has not yet been conducted. Here, we profiled the serum metabolomes of patients with three major forms of uveitis (Behҫet’s disease (BD), sarcoidosis, and Vogt-Koyanagi-Harada disease (VKH)) to identify potential biomarkers. This study included 19 BD, 20 sarcoidosis, and 15 VKH patients alongside 16 healthy control subjects. The metabolite concentrations in their sera were quantified using liquid chromatography with time-of-flight mass spectrometry. The discriminative abilities of quantified metabolites were evaluated by four comparisons: control vs. three diseases, and each disease vs. the other two diseases (such as sarcoidosis vs. BD + VKH). Among 78 quantified metabolites, 24 kinds of metabolites showed significant differences in these comparisons. Four multiple logistic regression models were developed and validated. The area under the receiver operating characteristic (ROC) curve (AUC) in the model to discriminate disease groups from control was 0.72. The AUC of the other models to discriminate sarcoidosis, BD, and VKH from the other two diseases were 0.84, 0.83, and 0.73, respectively. This study provides potential diagnostic abilities of sarcoidosis, BD, and VKH using routinely available serum samples that can be collected with minimal invasiveness.