Neural fingerprints of data driven cognitive subtypes across the psychosis spectrum: a B-SNIP study

Abstract

Cognitive dysfunction is a prominent feature of psychotic spectrum disorders. Identifying neurocognitive subgroups and their neural underpinnings may help elucidate distinct pathophysiological mechanisms and inform targeted interventions. This study aimed to derive cognitive subtypes using latent profile analysis (LPA) of the Brief Assessment of Cognition in Schizophrenia (BACS) and investigate associated variations in resting-state functional connectivity among these cognitive profiles and biologically derived Biotypes. The BACS was administered to 1807 psychosis patients from the B-SNIP1 and 2 cohorts to perform LPA and identify cognitive subgroups. Regional homogeneity (ReHo), a measure of local functional connectivity, was computed from resting-state fMRI data in a subset (717 patients, 427 controls). Multivariate regression models examined associations between ReHo and cognitive LPA, Biotypes, and DSM diagnostic categories. LPA identified four cognitive profiles: cognitively comparable to controls (CCC), intermediate-1, intermediate-2, and severely impaired. These profiles showed unique dysconnectivity patterns, particularly within the striatal, default mode, salience, and executive control networks. The severely impaired group exhibited hyperconnectivity in basal ganglia and executive control networks. The intermediate groups showed default mode and salience network connectivity disruptions. The CCC group was the least impaired, with hyperconnectivity in sensory and auditory networks. Compared to Biotypes, LPA subgroups presented more domain-specific connectivity fingerprints. Psychosis patients exhibit heterogeneous cognitive profiles with divergent intrinsic functional dysconnectivity patterns. Cognitive LPA subgroups demonstrated more domain-localized neural signatures than DSM subtypes, potentially allowing for more targeted interventions. This approach highlights the utility of cognitive subtyping using standardized cognitive assessments in elucidating pathophysiological mechanisms in psychosis.

Fig. 1

Detailed data processing pipeline.

Fig. 2

Cognitive performance for the four class-varying diagonal model.

Fig. 3. Cognitive performance across the various BACS subtests that were used for the latent profile analysis.

Line colors represent the four cognitive subgroups derived from the class-varying diagonal model. Depicted scores are Z-standardized and adjusted for age and sex. Significant between-proband differences in ReHo connectivity across A Cognitive B Biotype and C DSM subgroups. Curly brackets indicate significant effects at the p < 0.05 level, Sidak adjusted for multiple comparisons. Error bars refer to standard error of mean (SEM).