Supervised machine learning to decipher the complex associations between neuro-immune biomarkers and quality of life in schizophrenia

Abstract Stable phase schizophrenia is characterized by altered patterning in tryptophan catabolites (TRYCATs) and memory impairments, which are associated with PHEMN (psychosis, hostility, excitation, mannerism and negative) and DAPS (depression, anxiety and physio-somatic) symptoms. This study was carried out to examine the association between TRYCAT patterning, memory impairments, psychopathological features and health-related quality of life (HR-QoL) in schizophrenia. The World Health Organization (WHO) QoL instrument-Abbreviated version (WHO-QoL-BREF), IgA/IgM responses to TRYCATs, cognitive tests, Scale for the Assessment of Negative Symptoms (SANS), Hamilton and Depression (HAMD) and Anxiety (HAMA) Rating Scales and the Fibromyalgia and Chronic Fatigue Syndrome Rating Scale (FF) were measured in 80 schizophrenia patients and 40 controls. Neural Network analysis shows that the total HR-Qol score is best predicted by (in descending order) FF, HAMA and SANS scores, Mini Mental State examination, hostility, ratio noxious/protective TRYCATs and HAMD score. Partial least Squares (PLS) analysis shows that 55.1% of the variance in Domain1 (physical) is predicted by PHEMN and DAPS latent vector (LV) scores, while 57.9% of domain2 (psychological), 32.7% of domain3 (social) and 55.0% of domain4 (environment) are explained by DAPS LV scores. TRYCATs and episodic/semantic memory impairments have specific indirect effects on domains 2, 3 and 4, which are mediated by DAPS symptoms, while the effects of TRYCATs on domain1 are mediated by PHEMN and DAPS symptoms. Picolinic acid, xanthurenic acid and 3-hydroxy-kynurenine decrease WHO-QoL scores, whilst anthranilic acid is protective. The results show that lowered HR-Qol in schizophrenia is strongly predicted by noxious TRYCATs, impairments in episodic and semantic memory and DAPS symptoms, especially physio-somatic symptoms and anxiety. Neuro-immune pathways and the consequent cognitive impairments determine to a great extent lowered HR-QoL in schizophrenia.