A novel multi-class classification model for schizophrenia, bipolar disorder and healthy controls using comprehensive...

Area of Science:

• Neuroscience
• Genetics
• Psychiatry

Background:

• Schizophrenia (SCZ) and bipolar disorder (BP) are prevalent psychiatric disorders with overlapping symptoms and genetic risk factors.
• Current diagnostic methods rely on symptom-based assessments, highlighting the need for more objective diagnostic tools.
• The underlying etiology and pathology of SCZ and BP remain unclear, necessitating improved classification models.

Purpose of the Study:

• To develop a more accurate and consistent classification model for diagnosing schizophrenia (SCZ) and bipolar disorder (BP).
• To identify a gene signature that can differentiate between SCZ, BP, and healthy individuals.
• To assess the diagnostic capacity of a novel multi-class classification model.

Main Methods:

• Combined transcriptomic data from five independent studies, including 120 SCZ patients, 101 BP patients, and 149 healthy controls.
• Utilized partial least squares discriminant analysis (PLS-DA) to identify 341 differentially expressed genes (DEGs).
• Constructed a multi-class classification model using support vector machine (SVM) with the identified gene signature.

Main Results:

• Identified a significant gene signature comprising 341 DEGs with high disease relevance.
• Analyzed the protein-protein interaction network and gene ontology terms, revealing key roles in psychiatric disorders.
• The developed SVM multi-class model demonstrated strong classification ability on independent test sets.

Conclusions:

• The identified gene signature and the novel SVM-based classification model show promise for improving the diagnosis of SCZ and BP.
• This approach offers a potential pathway towards more objective and accurate diagnostic methods for these psychiatric conditions.
• Further validation is warranted to translate these findings into clinical practice.