Skip to main content
. 2021 May 10;17(5):e1008955. doi: 10.1371/journal.pcbi.1008955

Fig 4. Classification of MDD status and feature importance for the computational biomarker classifier.

Fig 4

Left, the receiver operating characteristic (ROC) curve and the ROC area under curve (AUC) statistic for a classifier using individual parameter values from the computational model (purple) and a classifier using mean observed behavioral measures of response time and approach rate and mean activity in ROIs (caudate nucleus, nucleus accumbens (NAcc), pregenual anterior cingulate cortex (pACC), and subthalamic nucleus (STN)). Right, mean estimated beta-coefficients from classifier with 95% confidence intervals for the classifier using computational biomarkers (purple). Coefficients are sorted by weight from left to right as the absolute distance from 0, the magnitude of which indicates the importance of each feature for the classification. Int = Intercept, PavBias = Pavlovian bias, v = drift rate, z = starting-point bias.