Discovering Genetic Variants in Hypertrophic Cardiomyopathy With Multiple Machine Learning Techniques.

Area of Science:

• Genetics
• Computational Biology
• Cardiology

Background:

• Hypertrophic cardiomyopathy (HCM) has significant genetic underpinnings.
• Understanding the complex interplay of genetic variants that modify HCM phenotype is crucial.
• High-dimensional genetic data presents challenges for traditional analysis.

Purpose of the Study:

• To analyze genetic variants in hypertrophic cardiomyopathy patients using diverse machine learning techniques.
• To identify relevant variants and understand their interactions.
• To discover potential disease modulators and pathogenic variants.

Main Methods:

• Statistical univariate analysis with p-value adjustment.
• Linear classifiers (SVM, FDA) for feature weighting.
• Informative variable identification and Bayesian networks for inter-variant relationships.
• Manifold learning for latent space representation.
• Linkage disequilibrium and frequency tables for variant association analysis.

Main Results:

• Ten genetic variants were consistently identified as significant across multiple methods.
• Twenty-two variants were significant in at least three out of five applied methods.
• Machine learning successfully detected disease-associated variants, including specific pathogenic founder variants.

Conclusions:

• Machine learning provides a robust framework for analyzing complex genetic data in hypertrophic cardiomyopathy.
• This methodology can identify significant disease-associated variants and potential genetic modulators.
• The findings contribute to a deeper understanding of the genetic architecture of HCM.