HAPSIM

HAPSIM generates case-control multi-locus genotype data and evaluates genotype–phenotype associations using artificial neural networks and permutation testing to localize disease loci without requiring haplotype phase or evolutionary modeling.

Key Features:

• Multi-Locus Haplotype Analysis: HAPSIM processes biallelic markers such as single nucleotide polymorphisms (SNPs) as multilocus haplotypes to provide information for localizing disease loci.
• Artificial Neural Network Utilization: It employs artificial neural networks (ANNs) to assess associations between marker haplotypes or multi-marker genotypes and disease phenotypes.
• Training and Output Generation: The ANN is trained on supplied marker genotypes and affection status and generates outputs approximating associated affection status for each genotype.
• Permutation Testing: HAPSIM implements permutation testing by randomizing genotypes relative to affection status to generate null datasets for evaluating statistical significance.
• Statistical Significance and Major Gene Effects: The method provides a robust measure of association between multi-marker genotypes and disease status, with particular sensitivity when major gene effects are present.
• Parameter Optimization: Network parameters are systematically varied to identify values that improve predictive accuracy.
• Conditional Analyses: The permutation framework supports conditional analyses by incorporating known risk factors alongside marker genotypes for independent assessment of marker contribution.
• Application Example: HAPSIM was applied to four SNPs in calpain 10 (CAPN10) in a case-control sample including type 2 diabetes, impaired glucose tolerance, and control subjects, yielding stronger evidence of association than single-marker tests adjusted for multiple testing.

Scientific Applications:

• Genetic Epidemiology: Investigation of multilocus genetic contributions to disease susceptibility in case-control studies.
• Complex Trait Analysis: Detection and assessment of multilocus associations and interactions in polygenic traits without requiring haplotype phasing.

Methodology:

HAPSIM uses artificial neural networks trained on marker genotypes and affection status to model genotype–disease relationships, generates predicted affection outputs per genotype, applies permutation testing by randomizing genotypes relative to affection status to evaluate significance, systematically varies network parameters for optimization, and supports conditional analyses by including known risk factors.