Extended Data Fig. 7: Predictions of models trained on discovery cohort on validation cohort.

a,b, Brain regions that can significantly predict subjective disgust revealed by searchlight- (Panel a) and parcellation-based (Panel b) analyses, respectively. Statistical significance was evaluated by prediction−outcome correlation (Pearson; two-sided; P < 0.001, uncorrected). Histograms: Predictions (correlations) from searchlights (Panel a) and parcellations (Panel b), respectively. The orange line indicates the prediction-outcome correlation from VIDS. c,d, Predictions (mean ± s.e.m.) from insula- (Panel c) and amygdala-based (Panel d) prediction analyses, respectively. Error bar indicates the s.e.m.; r indicates overall (between- and within-subjects; that is, n = 149 pairs) prediction-outcome Pearson correlation coefficient. e, The information about subjective experience of disgust is distributed across multiple systems. Model performance was evaluated as increasing numbers of voxels/features (x-axis) were used to predict subjective disgust in different regions of interest including the entire brain (black), consciousness network, subcortical regions or large-scale cerebral networks. The y-axis denotes the prediction-outcome correlation. Colored dots indicate the mean correlation coefficients, solid lines indicate the mean parametric fit and shaded regions indicate the s.d.