Identification of an inflammation-associated psychosis onset subgroup by applying unsupervised machine learning to whole-blood expression levels of immune gene transcripts

Abstract

Nowadays, the lack of quantitative criteria to resolve the diagnostic heterogeneity of psychotic onsets limits the development of safer and more effective treatments. Therefore, the hypothesis to integrate multimodal data to uncover biological subtypes of psychosis has risen [1,2]. Here we explored the existence of subgroups of patients affected by first episode psychosis (FEP) with a possible immunopathogenic basis.To do this, we designed a computational model that use unsupervised machine learning to cluster a sample of 127 FEP patients and 117 healthy controls (HC), based on the peripheral blood concentrations of 12 immune gene transcripts which demonstrated to classify with high accuracy between FEP patients and healthy subjects in a previous study [3]. To validate the model, we applied a resampling strategy based on the half-splitting of the total sample. Further, we tested the correlation between the subgroups and clinical, neuropsychological and brain structural variables.In both the discovery and validation samples, the model identified a FEP cluster characterized by the high expression of inflammatory and immune-activating genes (IL1b, CCR7 and IL12a) and of a single immune counterregulatory gene (CCR3)[4] and a further cluster consisting of equal number of FEP and HC subjects, which did not show a relative over or under expression of any immune marker (balanced subgroup). Also, none of the subgroups were related to specific symptoms dimensions or longitudinal diagnosis. FEP patients included in the balanced immune subgroup showed a reduced left hippocampal volume and a left supramarginal and lateroccipital cortexes’ thinning. These correlations seem to support an opposite pattern in the correspondent brain area of the inflammatory subgroup [5].Our results demonstrated the existence of a FEP patients’ subgroup that present a prominent activation of the inflammatory response. This evidence may pave the way to sample stratification in future trials aiming to develop personalized diagnostic tools and therapies targeting specific immunopathogenic pathways of psychosis onsets.