Searching for genotype-phenotype structure: using hierarchical log-linear models in Crohn disease

Area of Science:

• Genetics
• Computational Biology
• Medical Informatics

Background:

• Gene-disease associations are increasingly detected, but characterizing specific genetic variant effects requires advanced tools.
• Complex diseases like Crohn disease (CD) involve multiple subphenotypes, posing analytical challenges due to high inter-subphenotype correlations.
• Distinguishing causal genes for specific subphenotypes is difficult with traditional binary disease classifications.

Purpose of the Study:

• To develop and evaluate a novel analytical framework for detailed characterization of genetic variant effects on disease subphenotypes.
• To apply this framework to Crohn disease (CD) to uncover specific gene-subphenotype relationships.
• To improve the accuracy of genetic association studies by moving beyond simple binary disease indicators.

Main Methods:

• Utilized a model selection approach based on log-linear models within a Bayesian framework.
• Employed a reversible-jump Metropolis-Hastings algorithm for model fitting.
• Validated the method through simulation studies and application to real-world Crohn disease data.

Main Results:

• The developed method revealed a sparse disease structure in Crohn disease.
• The NOD2.908G-->R mutation was directly linked to more severe disease behaviors.
• Other NOD2 variants (1007L-->FS, 702R-->W) and the ATG16L1.300T-->A variant were associated with small bowel disease.

Conclusions:

• The proposed Bayesian framework effectively distinguishes gene associations with specific disease subphenotypes.
• This approach provides a more nuanced understanding of genetic contributions to complex diseases like Crohn disease.
• Specific genetic variants within NOD2 and ATG16L1 show distinct associations with disease severity and location in the gastrointestinal tract.