MAGprediction

MAGprediction predicts highly polymorphic gene alleles from unphased single nucleotide polymorphism (SNP) data to support interpretation of genome-wide association studies (GWAS).

Key Features:

• Allele prediction from unphased SNPs: Predicts highly polymorphic gene alleles using unphased single nucleotide polymorphism (SNP) data.
• Method foundation: Implements a complementary approach to the Identity-by-Descent (IBD) method originally proposed by Leslie et al. (2008).
• Training data: Uses unphased SNP data as the training dataset to build predictive models applicable to large population studies.
• HLA locus modeling: Constructs predictive models for HLA-A, HLA-B, HLA-C, HLA-DRB1, and HLA-DQB1.
• Cohort testing and validation: Models were trained on a cohort of 630 healthy individuals and validated on an independent cohort of the same size.
• Predictive accuracy: Achieved accuracies at intermediate or high resolution up to 100% for HLA-A, 98% for HLA-B, 98% for HLA-C, 97% for HLA-DRB1, and 98% for HLA-DQB1.
• Applicability beyond HLA: Demonstrates potential applicability to other highly variable genetic regions.
• Biological allele identification: Facilitates identification of functional gene alleles beyond SNP markers within or near genes to aid biological interpretation of associations.

Scientific Applications:

• GWAS interpretation: Supports interpretation and biological understanding of genome-wide association studies (GWAS) by linking allele variation to association signals.
• Complex-trait genetics: Enables investigation of genetic contributions to complex traits such as autoimmune diseases, infectious diseases, cancer, and heart disease.
• HLA-focused allele studies: Provides cohort-level HLA allele predictions to support allele–trait association analyses.

Methodology:

Implements a complementary approach to the Identity-by-Descent (IBD) method of Leslie et al. (2008), trains predictive models on unphased SNP data, and validates models on an independent cohort of equal size.