HAPAL

HAPAL models haplotype-haplotype interactions using penalized logistic regression with an adaptive L1-penalty to identify interacting haplotype blocks associated with complex disease risk.

Key Features:

• Adaptive Lasso Methodology: Employs an adaptive L1-penalty to perform simultaneous effect estimation and variable selection in haplotype interaction models.
• Penalized Logistic Regression Framework: Uses penalized logistic regression to model associations between haplotype interactions and binary outcomes.
• Modified Gauss-Seidel within EM Algorithm: Integrates a modified Gauss-Seidel method into the Expectation-Maximization (EM) algorithm for parameter estimation and selection.
• Higher-order Epistatic Interaction Handling: Designed to manage complexity arising from higher-order epistatic interactions among genetic variants.
• Performance Characteristics: Simulation studies report a low false positive rate and reasonable power for detecting significant haplotype interactions.
• Application Example: Applied to analyze mother and offspring genome interactions in small for gestational age (SGA) neonates.
• R Implementation: Implemented in R.

Scientific Applications:

• Genetic Association Studies: Identifying haplotype-haplotype interactions associated with disease susceptibility.
• Epistasis Investigation: Characterizing higher-order epistatic effects among multiple loci in complex diseases.
• Maternal-Offspring Interaction Analysis: Examining genome interactions between mothers and offspring, exemplified by studies of SGA neonates.
• Dissection of Genetic Architecture: Elucidating interacting haplotype block contributions to multifactorial disease etiology.

Methodology:

Implements penalized logistic regression with an adaptive L1-penalty (adaptive Lasso) to model haplotype-haplotype interactions and estimates parameters via a modified Gauss-Seidel method within an Expectation-Maximization (EM) algorithm.