Fig. 2
Association with disease and predictive accuracy of individual CSF proteins increases when adjusted for median CSF protein levels. (A) A linear model was constructed for each protein and disease, where the protein levels were the outcome variable. In the figure, the p-values for the diagnosis predictor (each disease vs. healthy control) are compared for models adjusted (y-axis) and unadjusted (x-axis) for median CSF protein levels. Only proteins with significant p-value in at least one of the models are included in the figure. The dashed line is drawn at p-value = 0.05. (B) ROC-AUCs resulting from logistic regression models predicting diagnosis (each disease vs. healthy controls) using the adjusted protein levels (y-axis) compared to unadjusted protein levels (x-axis). Only proteins with significant p-value in at least one of the models in (A) are included.
