Machine learning-based predictive models and subtypes patterns in peripheral blood of schizophrenia based on a...

Area of Science:

• Psychiatric Genetics
• Computational Biology
• Biomarker Discovery

Background:

• Schizophrenia (SCZ) pathogenesis remains poorly understood, hindering early diagnosis and effective treatment.
• Identifying reliable blood biomarkers and molecular subtypes for SCZ is a significant clinical challenge.

Purpose of the Study:

• To develop a robust diagnostic signature for SCZ using bioinformatics and machine learning.
• To identify distinct molecular subtypes of SCZ for personalized therapeutic strategies.

Main Methods:

• Integrated 12 machine learning algorithms across multiple SCZ datasets (GSE18312, GSE27383, GSE38485, GSE54913, GSE165604).
• Performed consensus clustering and non-negative matrix factorization (NMF) for subtype identification.
• Utilized GSEA, GSVA, Proteomaps, and IOBR analyses to characterize subtype differences.

Main Results:

• Identified a 16-gene diagnostic signature (APBB2, CLCN1, SYDE1, PAX5, SNAI1, DAZL, UNC93B1, PLAGL2, HS3ST1, ITPKB, PILRA, BTLA, SWAP70, AZI2, ADM, AVPR2) with high diagnostic performance across eight datasets.
• Developed a clinical nomogram for SCZ diagnosis and identified AZI2 as a key gene influencing inflammation and immunity.
• Discovered two distinct SCZ subtypes characterized by unique immune cell profiles and biological functions.

Conclusions:

• The study presents a validated 16-gene diagnostic signature and a novel nomogram for SCZ.
• The identified SCZ subtypes offer new insights into disease heterogeneity.
• Findings support the development of personalized diagnostic and treatment approaches for SCZ.