Figure 3.

Biosigner signature. (A) The algorithm assessed the relevance of the 160 metabolites identified for the prediction performances of Partial Least Squares-Discriminant Analysis (PLS-DA), Random Forest (RF), and Support Vector Machines (SVM) classifier models and subsequently identified 2 robust ‘S tier’ features, i.e., nicotinamide and N-acetyl L-leucine. The accuracies of the PLS-DA and SVM models on the final S signature were respectively 73.7% and 71.1% for nicotinamide and N-acetyl L-leucine. (B) The boxplots depict the metabolite levels in glaucoma and control groups. Error bars represent ± s.e.m and the solid bars within the boxplots represent the median level of metabolite (log transformed peak area) for each group.