Deficit schizophrenia is a discrete diagnostic category defined by neuro-immune and neurocognitive features: results of supervised machine learning

Buranee Kanchanatawan,Sira Sriswasdi,Supaksorn Thika,Sunee Sirivichayakul,André F. Carvalho,Michel Geffard,Marta Kubera,Michael Maes
DOI: https://doi.org/10.1007/s11011-018-0208-4
IF: 3.6551
2018-01-01
Metabolic Brain Disease
Abstract:Deficit schizophrenia is characterized by neurocognitive impairments and changes in the patterning of IgA/IgM responses to plasma tryptophan catabolites (TRYCATs). In the current study, supervised pattern recognition methods, including logistic regression analysis (LRA), Support Vector Machine (SVM), and Soft Independent Modeling of Class Analogy (SIMCA), were used to examine whether deficit schizophrenia is a discrete diagnostic class with respect to Consortium To Establish a Registry for Alzheimer’s disease (CERAD) and Cambridge Neuropsychological Test Automated Battery (CANTAB) tests and IgA/IgM responses to noxious (NOX) and generally more protective (PRO) TRYCATs. We recruited patients with ( n = 40) and without (n = 40) deficit schizophrenia and healthy volunteers (n = 40). The combined use of TRYCAT and CERAD features strongly segregates deficit from nondeficit schizophrenia and healthy controls. Three out of the top five most important features in LRA, SVM and SIMCA agreed, namely two different NOX/PRO TRYCAT ratios and false memory recall. SIMCA shows that deficit schizophrenia is significantly separated from nondeficit schizophrenia and controls with as top 6 features IgA responses to picolinic acid, IgM responses to 3-OH-kynurenine and kynurenic acid, and impairments in Word List Memory and Verbal Fluency Tests and Mini-Mental State Examination. Nevertheless, nondeficit schizophrenia was not significantly separated from controls. The results show that schizophrenia is not a unitary disease with mere continuous differences in severity of illness between apparent subtypes. Deficit schizophrenia is a qualitatively distinct class defined by neuroimmune (autoimmune responses to TRYCATs) and neurocognitive (episodic and semantic memory) features coupled or not with clinical (negative) symptoms.
What problem does this paper attempt to address?