Development and analysis of a comprehensive diagnostic model for aortic valve calcification using machine learning...

Area of Science:

• Cardiovascular Research
• Genomics and Bioinformatics
• Translational Medicine

Background:

• Advanced aortic valve calcification (AVC) requires timely diagnosis and intervention to prevent progression and improve patient outcomes.
• Current treatment options exist for symptomatic advanced AVC, but early detection remains crucial.
• Developing novel therapeutic targets is essential for better management of AVC.

Purpose of the Study:

• To identify potential diagnostic markers for early detection of aortic valve calcification (AVC).
• To develop a predictive model for AVC using machine learning and bioinformatics.
• To explore potential therapeutic targets for reversing AVC gene expression.

Main Methods:

• Utilized public expression profiles (GSE12644, GSE51472) to identify differentially expressed genes (DEGs) in AVC.
• Applied R software, random forest, support vector machines, and artificial neural networks for marker screening and modeling.
• Performed CIBERSORT immune infiltration analysis and utilized the CMap database for drug prediction.

Main Results:

• Identified 78 DEGs, with leukocyte migration and pid integrin 1 pathways being highly enriched.
• Validated CXCL16, GPM6A, BEX2, S100A9, and SCARA5 as significant diagnostic markers for AVC.
• Developed a molecular diagnostic score system with high accuracy (AUC=0.987) and identified Doxazosin and Terfenadine as potential therapeutics.

Conclusions:

• CXCL16, GPM6A, BEX2, S100A9, and SCARA5 are promising for AVC diagnosis and treatment.
• A machine learning-based molecular prognostic score system demonstrates significant diagnostic value for AVC.
• Immune cell infiltration, including B cells and macrophages, plays a role in AVC development and incidence.