Symbolic modeling of epistasis

Area of Science:

• Genetics
• Bioinformatics
• Computational Biology

Background:

• Parametric linear models are standard in genetic analysis but assume data linearity.
• Complex biological systems, like disease susceptibility, often exhibit nonlinear genotype-phenotype relationships due to epistasis and other factors.
• Existing methods struggle with these nonlinearities, limiting accurate genetic analysis.

Purpose of the Study:

• To introduce a novel five-step process for symbolic model discovery using Symbolic Discriminant Analysis (SDA).
• To address the limitations of linear models in capturing complex, nonlinear genotype-phenotype relationships.
• To enhance the interpretation of complex genetic models through function mapping and interaction information.

Main Methods:

• Symbolic Discriminant Analysis (SDA) for flexible, assumption-free data modeling.
• Integration of genetic programming (GP), experimental design, graphical modeling, and estimation of distribution algorithms.
• Development of function mapping for interpreting symbolic discriminant functions and interaction information.

Main Results:

• A robust five-step process for symbolic model discovery is presented.
• Function mapping combined with interaction information provides graphical decomposition of complex models.
• The approach effectively highlights synergistic, redundant, and independent effects of genetic polymorphisms.

Conclusions:

• The new SDA modeling process offers a powerful alternative for analyzing complex genetic data, especially in disease susceptibility.
• This method overcomes the restrictive assumptions of linear models.
• It facilitates deeper statistical interpretation of genotype-phenotype relationships in complex biological systems.