Machine learning-based integration identifies a 10-gene predictive signature and its classification patterns in...

Area of Science:

• Genetics and Genomics
• Computational Biology
• Psychiatric Disorders

Background:

• Schizophrenia (SCZ) is a complex psychiatric disorder with high heritability but unclear etiology.
• Current diagnostic and treatment biomarkers for SCZ are limited, hindering effective predictive, preventive, and personalized medicine (PPPM/3PM).

Purpose of the Study:

• To identify novel biomarkers for schizophrenia (SCZ) diagnosis and treatment.
• To develop a robust diagnostic signature and predictive model for SCZ using machine learning.
• To investigate SCZ subtypes for tailored 3PM strategies.

Main Methods:

• Differential gene expression (DEGs) and weighted gene co-expression network (WGCNA) analyses were performed on brain datasets.
• A machine learning (ML) framework with 12 MLs and 84 combinations was used to construct a consensus diagnostic signature.
• Consensus clustering and non-negative matrix factorization (NMF) were applied to identify SCZ subtypes.

Main Results:

• 53 SCZ-key genes were identified, leading to a consensus diagnostic signature with high discriminative performance.
• A nomogram model was established for quantitative SCZ prediction in clinical practice.
• SCZ patients were classified into two distinct subtypes with unique immune and metabolic profiles.

Conclusions:

• A novel diagnostic signature and nomogram were developed, offering high accuracy for SCZ diagnosis.
• Identified SCZ subtypes exhibit distinct inflammatory, immune, and metabolic patterns.
• Integrating SCZ subtypes into the 3PM framework presents opportunities for enhanced clinical intelligence and management.