Machine learning-based microarray analyses indicate low-expression genes might collectively influence PAH disease.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Pulmonary arterial hypertension (PAH) requires accurate subtype prediction for targeted therapy.
• Current methods may lack cost-effectiveness or early-stage diagnostic precision.
• Understanding PAH's genetic basis is crucial for developing novel treatments.

Purpose of the Study:

• To develop a cost-effective method for predicting PAH subtypes using microarray gene expression data.
• To identify a minimal set of highly informative genes for accurate PAH classification.
• To enhance the understanding of molecular pathways involved in PAH etiology.

Main Methods:

• Utilized a high-quality patient PAH dataset with microarray expression data.
• Applied various gene filtering techniques for initial data processing.
• Developed and implemented a novel feature selection and refinement algorithm.
• Integrated machine learning methods for predictive model construction.

Main Results:

• Clusters of low-expression genes proved highly informative for predicting and differentiating PAH forms.
• The novel feature refinement algorithm significantly improved model performance.
• Near-perfect classification accuracies were achieved using a small set of genes (around ten).

Conclusions:

• Machine learning models, combined with novel feature refinement, can accurately classify PAH subtypes.
• Cost-effective microarray data and a small gene set offer a promising approach for early PAH diagnosis and management.
• This study identifies key genes, advancing the understanding of PAH pathogenesis and molecular mechanisms.