An Efficient Nonlinear Regression Approach for Genome-wide Detection of Marginal and Interacting Genetic Variations

Area of Science:

• Genetics
• Statistical genomics
• Bioinformatics

Background:

• Genome-wide association studies (GWAS) identify genetic variants linked to traits.
• Detecting pairwise genetic variant interactions is computationally challenging due to vast combinations and small sample sizes.
• Existing methods struggle with scalability, distinguishing true signals, detecting nonlinear effects, and controlling false positives.

Purpose of the Study:

• To develop a unified framework (SPHINX) for efficient and accurate detection of pairwise genetic variant interactions.
• To address challenges including computational complexity, signal-to-noise ratio, nonlinear associations, and false positive control.
• To identify novel genetic associations with complex traits, such as late-onset Alzheimer's disease.

Main Methods:

• Proposed a piecewise linear model for interaction detection, balancing simplicity and ability to capture nonlinear effects.
• Introduced randomized group lasso with stability selection for statistical inference.
• Developed a screening algorithm to handle large-scale SNP pair analysis and computational challenges.

Main Results:

• SPHINX demonstrated superior power for interaction detection compared to existing methods while maintaining false positive control.
• Applied to a late-onset Alzheimer's disease dataset, SPHINX identified 16 single nucleotide polymorphisms (SNPs) and 17 SNP pairs associated with disease traits.
• A scalable implementation screened approximately 118 billion association candidates in under 5.5 hours on a 60-node cluster.

Conclusions:

• SPHINX offers a robust and scalable solution for detecting genetic variant interactions, overcoming limitations of previous approaches.
• The framework effectively identifies significant genetic associations, including complex nonlinear interactions, in large-scale genomic datasets.
• SPHINX provides a powerful tool for genetic research, particularly in complex diseases like Alzheimer's, accelerating discovery and improving biological understanding.