Fig. 1: Flowchart for developing, testing and validating the SZ PRS-associated multimodal pattern.

a SZ PRS was used as a reference to guide fALFF + GMV fusion to identify PRS-associated multimodal networks. b PRS-associated networks were separated as positive (Z > 0) and negative (Z < 0) brain regions based on the Z-scored brain maps, resulting in 4 features for the following analysis. c The same fusion with PRS analysis was performed on healthy white people, healthy subjects, and all the available subjects within UKB with PRS thresholded at P_SNP = 5.0e−08, 1.0e−04 and 0.05, and pruned at r² < 0.1 and 0.2, respectively. d Group differences between SZ and HC of the identified PRS-associated features were tested across 4 independent SZ cohorts (including BSNIP-1, COBRE, fBIRN and MPRC). e Linear SVM was used to classify SZ and HC across 4 independent SZ cohorts based on PRS-associated features (including BSNIP-1, COBRE, fBIRN and MPRC); multiple linear regression was performed to predict cognition and symptom for SZ across 3 independent cohorts (BSNIP-1, COBRE, and fBIRN). MPRC data were not included in the prediction analysis since the related clinical measures were not available. f Group differences between SZ and other psychosis subjects and their relatives, and between other psychiatric disorders (ASD, MDD, ADHD) and HC were tested. UKB (UK Biobank, N = 37,347); BSNIP-1 (Bipolar and Schizophrenia Network for Intermediate Phenotypes), SZ: N = 178; HC: N = 220; COBRE (Center for Biomedical Research Excellence), SZ: N = 100; HC: N = 90; fBIRN (Functional Imaging Biomedical Informatics Research Network), SZ: N = 164; HC: N = 157; MPRC (Maryland Psychiatric Research Center), SZ: N = 164; HC: N = 157; ASD autism spectrum disorder, MDD major depressive disorder, ADHD attention-deficit/hyperactivity disorder, SVM support vector machine.