Alternating optimization for G × E modelling with weighted genetic and environmental scores: Examples from the MAVAN...

Area of Science:

• Genetics
• Environmental Health
• Biostatistics

Background:

• Existing Genotype × Environment (G × E) models often struggle to incorporate multiple genetic variants and environmental exposures simultaneously in a parsimonious manner.
• There is a need for methods that can handle complex interactions while maintaining model simplicity.

Purpose of the Study:

• To develop a novel method for estimating parameters in G × E models that can simultaneously include multiple genetic variants and environmental exposures.
• To create a parsimonious G × E model structure that accommodates complexity.

Main Methods:

• Developed a novel method using alternating optimization to estimate parameters in G × E models.
• The genetic score (G) and environmental score (E) weights, along with main model parameters, are estimated iteratively.
• The model is implemented as a 2-way interaction longitudinal mixed model, extendable to k-way interactions and generalized linear mixed models.

Main Results:

• The alternating optimization approach allows for the construction of complex interaction models with constrained structures and fewer parameters.
• Simulations demonstrate the power and validity of the approach, even with small sample sizes.
• Application to the Maternal Adversity, Vulnerability, and Neurodevelopment (MAVAN) study showed significant improvements over existing models.

Conclusions:

• The novel method provides a powerful and valid approach for analyzing complex G × E interactions.
• The implemented R (LEGIT) and SAS (LEGIT_SAS) packages facilitate the application of this method.
• This technique enhances the ability to model intricate relationships between genetic and environmental factors in various research settings.