STATISTICAL GENOMIC MODELS FOR IDENTIFYING RARE VARIANTS ASSOCIATED WITH COMPLEX HUMAN DISEASES

Abstract

Background: Rare genetic variants are known to play an important role in the development of complex human diseases such as cardiovascular diseases, diabetes mellitus, neurodegenerative diseases and cancer. Conventional genome-wide association studies (GWAS) typically have limited sensitivity for low frequency variants associated with multifactorial diseases, necessitating advanced statistical genomic approaches.

Objective: This study assesses statistical genomic models for identifying rare variant associations with complex human diseases and compares the performance of statistical and machine learning-based genomic analyses.

Methods: Comparative genomic analysis was performed on datasets from UK Biobank, 1000 Genomes Project, and dbGaP repositories. Statistical methods (GWAS, Sequence Kernel Association Test (SKAT), burden analysis, deep learning genomic models) were used to detect rare variants and to predict disease risk.
Findings: Deep learning models showed the highest rare variant detection accuracy of 91%. For gene-level association analysis, SKAT showed 85% sensitivity. The integrated statistical and machine learning approaches improved disease risk prediction by approximately 38% over traditional GWAS methods. Further functional annotation analysis revealed several pathogenic variants associated with cardiovascular and neurodegenerative disorder