Development and validation of preeclampsia predictive models using key genes from bioinformatics and machine learning...

Area of Science:

• Reproductive Biology
• Immunology
• Genetics

Background:

• Preeclampsia (PE) presents significant diagnostic and therapeutic challenges.
• Identifying novel diagnostic and therapeutic targets is crucial for managing PE.
• Understanding the immune mechanisms underlying PE is essential for developing effective treatments.

Purpose of the Study:

• To identify novel genes for potential diagnostic and therapeutic targets in preeclampsia (PE).
• To elucidate the immune mechanisms involved in PE pathogenesis.
• To develop predictive models for PE diagnosis and prognosis.

Main Methods:

• Analysis of three GEO datasets, including differential gene expression and Weighted Gene Co-expression Network Analysis (WGCNA).
• Application of machine learning algorithms (LASSO, SVM-RFE, RF) to identify diagnostic hub genes.
• Development of a predictive nomogram and a Fully Connected Neural Network (FCNN) for PE prediction, alongside immune infiltration analysis (ssGSEA).

Main Results:

• Five validated diagnostic biomarkers for PE were identified: CGB5, LEP, LRRC1, PAPPA2, and SLC20A1, with high AUC values.
• A predictive nomogram demonstrated strong predictive power (C-index 0.873), and FCNN achieved an AUC of 0.911.
• Immune infiltration analysis highlighted the involvement of T cell subsets, neutrophils, and NK cells in PE, linking identified genes to immune responses.

Conclusions:

• CGB5, LEP, LRRC1, PAPPA2, and SLC20A1 are validated as key diagnostic biomarkers for preeclampsia (PE).
• The developed nomogram and FCNN models offer credible prediction of PE.
• The association of these biomarkers with immune cell infiltration underscores the critical role of immune responses in PE pathogenesis.