Construction and evaluation of a metabolic correlation diagnostic model for diabetes based on machine learning...

Area of Science:

• Genomics
• Molecular Biology
• Bioinformatics

Background:

• Diabetes mellitus (DM) is a widespread metabolic disorder requiring better understanding of its molecular underpinnings for improved diagnosis and treatment.
• Identifying specific molecular markers is crucial for early detection and therapeutic intervention in DM.

Purpose of the Study:

• To develop a robust diagnostic model for diabetes mellitus using gene expression data.
• To identify key genes with diagnostic value for DM.
• To explore the potential for subclassification of DM patients based on molecular profiles.

Main Methods:

• Utilized public gene expression datasets (GSE7014, GSE25724, GSE156248) for model development and (GSE20966) for validation.
• Employed Random Forest and LASSO regression for building a diagnostic classifier.
• Performed functional enrichment analysis and validated key gene expression (CXCL12, PPP1R12B) via qRT-PCR in PBMCs.

Main Results:

• Identified 131 differentially expressed genes (DEGs) between DM patients and controls.
• Discovered significant pathway alterations including IL-12 signaling, JAK signaling, and ferroptosis in DM.
• Validated five genes (CXCL12, MXRA5, UCHL1, PPP1R12B, C7) with high diagnostic accuracy, confirmed by qRT-PCR showing CXCL12 and PPP1R12B upregulation in DM patients.

Conclusions:

• Developed a highly accurate diagnostic model for DM based on a panel of five key genes.
• Demonstrated the utility of the gene set for subtyping DM patients, revealing distinct functional characteristics.
• Provided novel insights into the molecular mechanisms and pathogenesis of diabetes mellitus.