Identifying neurobiological heterogeneity in clinical high-risk psychosis: a data-driven biotyping approach using resting-state functional connectivity.

Abstract

To explore the neurobiological heterogeneity within the Clinical High-Risk (CHR) for psychosis population, this study aimed to identify and characterize distinct neurobiological biotypes within CHR using features from resting-state functional networks. A total of 239 participants from the Shanghai At Risk for Psychosis (SHARP) program were enrolled, consisting of 151 CHR individuals and 88 matched healthy controls (HCs). Functional connectivity (FC) features that were correlated with symptom severity were subjected to the single-cell interpretation through multikernel learning (SIMLR) algorithm in order to identify latent homogeneous subgroups. The cognitive function, clinical symptoms, FC patterns, and correlation with neurotransmitter systems of biotype profiles were compared. Three distinct CHR biotypes were identified based on 646 significant ROI-ROI connectivity features, comprising 29.8%, 19.2%, and 51.0% of the CHR sample, respectively. Despite the absence of overall FC differences between CHR and HC groups, each CHR biotype demonstrated unique FC abnormalities. Biotype 1 displayed augmented somatomotor connection, Biotype 2 shown compromised working memory with heightened subcortical and network-specific connectivity, and Biotype 3, characterized by significant negative symptoms, revealed extensive connectivity reductions along with increased limbic-subcortical connectivity. The neurotransmitter correlates differed across biotypes. Biotype 2 revealed an inverse trend to Biotype 3, as increased neurotransmitter concentrations improved functional connectivity in Biotype 2 but reduced it in Biotype 3. The identification of CHR biotypes provides compelling evidence for the early manifestation of heterogeneity within the psychosis spectrum, suggesting that distinct pathophysiological mechanisms may underlie these subgroups.