Extended Data Fig. 3: Association between the discriminative FCs and treatment-outcome-predictive FCs.

a,b, Correlation between the top 2 active rTMS treatment response predictive FCs and active rTMS VAS score change. These two FCs were between the orbitofrontalcortex and anterior cingulate cortex, and between the middle temporal cortex and superior orbitofrontalcortex. Error bars = s.e.m. represent the Pearson’s correlation between FCs and VAS score changes based on the two-sided test against the alternative hypothesis that r ≠ 0. c,d, Correlation between these two FCs and sham rTMS VAS score change. e,f, These two FCs distribution between CUD and HC in the discovery and independent cohorts. The difference of FCs were comfirmed by independent samples t-test (two-sided test against the alternative hypothesis that t ≠ 0). The data in discovery cohort was augmented twice. These two FCs were significantly and specifically correlated to the VAS score change and significantly different between CUD and HC. The boxplots show the interquartile range (IQR; first quartile, 25th percentile; third quartile, 75th percentile), and the whiskers indicate Q1 − (1.5 × IQR) or Q3 + (1.5 × IQR). The line within the boxplot represents the median. The sample sizes for all panels were n = 213 (augmented FC from 71 subjects) for the CUD in discovery cohort, n = 174 (augmented FC from 58 subjects) for the HC in discovery cohort, n = 82 for the CUD in independent cohort, n = 81 for the HC in independent cohort. g, Venn diagram indicating the association between discriminative and abnormal FCs (551) with active rTMS treatment outcome. Discriminative atypical FCs were defined as the discriminative FCs identified by our classification models and the significantly atypical FCs detected by two-sample t-tests (two-sided test against the alternative hypothesis that t ≠ 0) comparing CUD and HC subjects, with those surviving FDR correction (pfdr < 0.05). The number of discriminative atypical FCs was equal to the sum of hyperconnections and hypoconnections. Deeper bluer shading indicates larger treatment predictive weights. The red numbers in the red rectangle represent the overlapping numbers between the top 100 treatment predictive FCs and all discriminative atypical FCs in descending order.