Fig. 1: Analysis overview.

a The data utilised is from two existing datasets: CamCAN ageing and UCLA phenomics, the latter including healthy individuals as well as data from three clinical cohorts (Schizophrenia - SCHZ, bipolar disorder - BPD, and ADHD). b 28 different symptom scores available transdiagnostically were used to examine within- and between-group similarity in terms of symptomatology (I) as well as derive composite symptom sub-domains using PCA (ii). c REACT was used to generate molecular-enriched functional networks (i) which were subsequently parcellated using a custom combination of cortical, subcortical, and cerebellar ROIs (ii). d These ROI-based molecular-enriched networks were then used to create normative models trained on 70% of the HC subjects from both CamCAN and UCLA (i), then used to characterise deviations from normality within the remaining 30% of HC as well as the three clinical cohorts (ii). e Deviation scores within each brain region and averaged across brain regions as summary metrics were compared using ANOVAs (i); the averaged summary metrics were also examined for classification value using binary logistic regression (ii); within- and between-group FC deviation similarity was also evaluated (iii). f Finally, deviation scores were analysed transdiagnostically, examining how transdiagnostic similarity relates to the extent of deviation and composite symptom sub-domains identified in B (i). Regional deviations were also correlated with symptom sub-domains (ii).