Multifactor dimensionality reduction as a filter-based approach for genome wide association studies.

Area of Science:

• Genetics
• Bioinformatics
• Statistical genomics

Background:

• Advances in genotyping and genetic data facilitate understanding human genetic diseases via variation studies.
• Genome-wide association studies (GWAS) traditionally identify univariate associations, but epistatic (interaction) effects are increasingly recognized.
• Prioritizing significant findings from GWAS for replication and functional studies is crucial, necessitating effective filtering approaches.

Purpose of the Study:

• To evaluate multifactor dimensionality reduction (MDR) as a filter for prioritizing genetic variants in GWAS.
• To assess MDR's ability to detect both univariate and epistatic genetic effects.
• To compare MDR's performance against logistic regression (LR) and evaporative cooling (EC) filtering methods.

Main Methods:

• Simulated genetic data with diverse effect sizes were used to evaluate MDR.
• MDR was compared with logistic regression (LR) and evaporative cooling (EC) for variant prioritization.
• Performance was assessed for detecting main genetic effects and two-locus epistatic interactions.

Main Results:

• MDR effectively prioritized variants, detecting main effects and interactions with or without marginal effects.
• MDR performed comparably to EC and LR for main effect models.
• MDR significantly outperformed LR in detecting two-locus epistatic models, showing equivalent performance to EC.

Conclusions:

• Multifactor dimensionality reduction (MDR) is a viable and effective filter for prioritizing genetic variants in GWAS.
• MDR demonstrates strong capability in detecting gene-gene interactions (epistasis), a critical aspect often missed by traditional methods.
• The study highlights MDR's potential to enhance the discovery of complex genetic architectures underlying human diseases.