Extended Data Fig. 9: Additional univariate and multivariate analyses between disease groups.

a, The univariate Cohen’s d distribution of global and regional brain structural measures across diseases groups. b, A summary of the global and regional deviation scores showing largest Cohen’s d and corresponding overlapping coefficient and classification performance between disease groups and HCs. c, The univariate Cohen’s d between each paired disease groups. * pFDR<0.05, ** pFDR<0.005, *** pFDR<0.001. d, Multivariate classification between disease groups. SVM with 10-fold cross-validation was conducted, similar to the DPS model development. e, The contributing measures of the disease-specific classification models for DPS estimation. For each disease, the measures that selected in all the SVM 10-folds were defined as contributing measures, and they were ordered by their univariate Cohen’s d (HCs versus disease) in a decreasing manner. f, The confusion matrix of the disease-specific classification models for each disease. g, The distribution of DPS across diseases by using disease-specific classification models. ANOVA was used for statistical analysis. h, The distributions of 1000 permuted Pearson’s correlation and MAPE for clinical score and PD outcome prediction. j, The PD outcome prediction using baseline UPDRS alone. k, The univariate Pearson’s correlation between the deviation scores clinical scores across diseases. Note: MCI, mild cognitive impairment; AD, Alzheimer’s disease; PD, Parkinson’s disease; CSVD, cerebral small vessel disease; MS, multiple sclerosis; NMOSD, neuromyelitis optica spectrum disorder; AUC, area under the curve; GMV, gray matter volume; sGMV, subcortical gray matter volume; WMV, white matter volume; CSF, cerebrospinal fluid; SVM, support vector machine; DPS, disease propensity score; MAPE, mean absolute percentage error; MMSE, Mini-Mental State Examination; MoCA, Montreal Cognitive Assessment; UPDRS, Unified Parkinson’s Disease Rating Scale; EDSS, Expanded Disability Status Scale; DBS, deep brain stimulation; FDR, false discovery rate. All statistical tests were two-sided with FDR correction for multiple comparisons (if relevant).