According to the Vision Health Initiative Project, glaucoma, a common eye disease that damages the optic nerve, is the second leading cause of blindness worldwide. Glaucoma, if treated at its initial stages, allows one to avoid permanent vision impairment. The problem arises because, at early stages, there are no visible symptoms, which prevent correct diagnostics. Our goal is to develop reliable methods to diagnose the early stages of glaucoma using ‘omics data and machine learning (ML). Collecting metabolomic data from public sources—articles and databases, we preprocessed them, calculated descriptors, and then fed the data into different WEKA’s algorithms, creating the models and training and testing them. We also created datasets of random metabolites independent of glaucoma. The testing of several models shows that the best is the Hoeffding tree model, revealing an accuracy of 90.91%. Testing the model on an independent dataset yielded an accuracy of 79.24%. A developed ML model could feasibly detect early-onset glaucoma using a patient’s metabolomic profile. Future studies with a larger pool of metabolite data could improve the accuracy of diagnostics.