Evaluating GWAS Model Performance Across Heritability and Polygenicity Gradients in Simulated Plant Trait Data

Genome-wide association studies (GWAS) are a widely applied approach for dissecting complex traits in crops. In this study, we systematically compared the performance of eight GWAS models across a range of heritability levels (0.3–0.8), followed by deeper analysis of the top three models—BLINK, FarmCPU, and MLMM—under two polygenic scenarios involving 50 and 100 quantitative trait loci (QTLs). Our results highlight the importance of trait heritability and genetic architecture in determining GWAS model effectiveness. BLINK consistently detected the highest number of true positives (TPs), particularly for moderately heritable traits, while MLMM showed superior mapping resolution. These findings provide practical guidance for selecting appropriate GWAS methods based on trait complexity.