Related Experiment Video
Updated: Aug 27, 2026
Handwriting Analysis Indicates Spontaneous Dyskinesias in Neuroleptic Naïve Adolescents at High Risk for Psychosis
Published on: November 21, 2013
Early-life inflammatory subgroups in youth with psychotic experiences: A semi-supervised machine learning approach
Shalini Bhattacharya1, Kate Merritt2, Golam M Khandaker3
1Division of Psychiatry, University College London, 4th Floor, Maple House, 149 Tottenham Court Road, London W1T 7BN, UK; Department of Clinical, Educational and Health Psychology, University College London, 4(th) Floor, 26 Bedford Way, London WC1H 0AP, UK.
Aims:
Psychotic experiences (PEs) are relatively common in youth and are associated with increased risk for later psychotic disorders. Although inflammation is implicated in the pathophysiology of psychosis, it remains unclear whether specific immune signatures are present before the emergence of PEs. This study aims to identify inflammatory subgroups among individuals with PEs in early adulthood, based on childhood inflammatory markers.
Methods:
Data on 38 immune markers at age nine and PEs at age 18 in individuals from the Avon Longitudinal Study of Parents and Children (ALSPAC) birth cohort were analysed. The sample comprised 2,189 participants, including 156 individuals with PEs and 2,033 healthy controls without PEs. HYDRA (Heterogeneity through Discriminative Analysis), a semi-supervised machine learning approach, was used to identify subgroups within those with PEs while using healthy controls without PEs as a reference. We then compared these subgroups to determine whether they differed in PE severity and in the diagnosis of depression.
Results:
HYDRA identified two subgroups of young adults with PEs, one with and one without evidence of immune activation in childhood (Adjusted Rand Index = 0.47). The immune activation subgroup was characterised by higher levels of inflammatory markers (28 out of 38 upregulated), including cytokines, chemokines, and matrix metalloproteinases, compared to the other PE group and compared to healthy controls without PEs. Increasing PE severity (suspected PE, definite PE, and psychotic disorder) was associated with increased odds of belonging to the immune activation subgroup (OR = 1.63, 95% CI 1.04-2.58, p = 0.034), whereas there were no differences in depression diagnoses between subgroups (OR = 0.81, 95% CI 0.35-1.86, p = 0.633).
Conclusion:
These findings provide the first evidence that childhood inflammatory subgroups, defined by a broad panel of immune markers, may be linked to the later emergence of PEs in adolescence and early adulthood. Using a semi-supervised, data‑driven machine learning approach, this study identifies early-life immune markers that may indicate vulnerability pathways to psychosis.
