A machine learning-based SNP-set analysis approach for identifying disease-associated susceptibility loci

Area of Science:

• Genetics
• Computational Biology
• Immunology

Background:

• Identifying disease susceptibility loci is crucial for understanding complex diseases.
• Current methods for biomarker discovery have limitations, including underpowered detection and neglecting variant interactions.
• The "missing heritability" problem requires innovative approaches to genetic association studies.

Purpose of the Study:

• To discover disease-associated susceptibility loci by enhancing genome-wide association studies (GWAS).
• To integrate random forest and cluster analysis for improved biomarker discovery.
• To identify novel genetic loci associated with hepatitis B virus surface antigen (HBsAg) seroclearance.

Main Methods:

• Augmented a GWAS using an integrated framework of random forest and cluster analysis.
• Performed cluster analyses on significant single nucleotide polymorphisms (SNPs) from GWAS and SNPs with high feature importance from random forest.
• Tested resulting SNP-sets for trait association with HBsAg seroclearance.

Main Results:

• Identified three potential susceptibility loci associated with HBsAg seroclearance: SNP rs2399971, gene LINC00578, and locus 11p15.
• SNP rs2399971 was previously reported as significantly associated with HBsAg seroclearance in treated patients.
• LINC00578 and 11p15 are linked to diseases affected by hepatitis B virus infection.

Conclusions:

• The integrated framework shows potential for identifying disease-associated susceptibility loci.
• Findings may improve understanding of complex disease etiologies.
• Results could inform advanced disease risk assessment for patients.