Comparing mixed models and Random Forest association tests using naturalgwas and a Striped Bass SNP data set.

Area of Science:

• Genetics
• Bioinformatics
• Statistical Genomics

Background:

• Genome-Wide Association Studies (GWAS) are crucial for identifying genetic variants associated with traits.
• Accurate statistical methods are essential to control for false discoveries and genomic inflation in GWAS.
• Evaluating new methods against established ones is vital for advancing genetic research.

Purpose of the Study:

• To assess the performance of Zhao's Random Forest method for genotype-phenotype association testing.
• To compare Zhao's Random Forest against uncorrected Random Forest, LFMM, GEMMA, and CATE.
• To evaluate association tests' power, false discovery rates, and false positives using simulated data.

Main Methods:

• Phenotype simulation using the naturalgwas package.
• Simulation of 400 phenotype sets across varying effect sizes and causal loci (2-30).
• Evaluation of five association methods: Zhao's Random Forest, uncorrected Random Forest, LFMM, GEMMA, and CATE.

Main Results:

• Zhao's Random Forest and GEMMA exhibited the lowest genomic inflation.
• All tested methods showed similar power in detecting causal loci.
• Zhao's Random Forest consistently yielded the lowest false discovery rate and fewest false positives across all scenarios.

Conclusions:

• Zhao's Random Forest is a robust method for genotype-phenotype association testing, outperforming other methods in controlling false discoveries.
• The simulation approach provides a valuable framework for assessing GWAS method performance on empirical data.
• This evaluation aids in selecting appropriate statistical tools for future GWAS.