Identification of multiple gene-gene interactions for ordinal phenotypes

Area of Science:

• Genetics
• Biostatistics
• Computational Biology

Background:

• Multifactor Dimensionality Reduction (MDR) is effective for gene-gene interaction analysis in complex diseases.
• Traditional MDR is limited to binary traits, while many traits, like obesity, are ordinal.
• There is a need for methods to analyze gene-gene interactions in studies with ordinal traits.

Purpose of the Study:

• To propose Ordinal Multifactor Dimensionality Reduction (OMDR) for gene-gene interaction analysis in ordinal traits.
• To introduce tau-b as an association measure and Generalized Cross-Validation Consistency (GCVC) for evaluating interactions.
• To demonstrate the utility of OMDR and GCVC in genetic studies with ordinal outcomes.

Main Methods:

• Developed Ordinal Multifactor Dimensionality Reduction (OMDR) to handle ordinal traits.
• Utilized tau-b, an ordinal association measure, for interaction evaluation.
• Generalized Cross-Validation Consistency (GCVC) was employed to identify multiple optimal gene-gene interactions.

Main Results:

• OMDR demonstrated superior performance over MDR in simulations, showing improved power, predictability, and selection stability.
• Analysis of real body mass index (BMI) data revealed more interactions for ordinal traits compared to binary traits.
• Interactions identified using OMDR for ordinal traits exhibited higher predictability on average than those from MDR for binary traits.

Conclusions:

• OMDR is a powerful and effective method for gene-gene interaction analysis in studies with ordinal traits.
• The proposed OMDR and GCVC methods offer practical advantages for analyzing complex traits, especially in large genetic studies.
• Software implementing OMDR and GCVC is available for non-commercial research.