Systematic Comparison of GWAS Methods in Wheat: Balancing Statistical Power, False Positive Control, and Computational Efficiency

Genome-wide association studies (GWAS) are essential for identifying genetic loci associated with complex traits, but the choice of statistical method significantly influences performance. This study systematically compared eight GWAS methods (GLM, MLM, CMLM, SUPER, MLMM, ECMLM, FarmCPU, and BLINK) in wheat, using simulated phenotypic data across 12 replicates to assess statistical power, false discovery rate (FDR), Type I error control, and computational efficiency. Phenotypes were simulated with a heritability of 0.7 and 10 quantitative trait nucleotides (QTNs) using genotypic data from 110 wheat accessions with 5,587 SNPs. Results indicated that multi-locus methods, particularly MLMM, FarmCPU, and BLINK, outperformed single-locus approaches, achieving higher power at lower FDR thresholds (MLMM: 0.434 at FDR = 0.0003; FarmCPU: 0.309 at FDR = 0.0007) and better Type I error control (MLMM: 0.943 at Type I error = 0.033; BLINK: 0.749 at Type I error = 0.073). MLMM exhibited the highest area under the curve (AUC) for power versus FDR (0.164 ± 0.105) and Type I error (0.823 ± 0.045), while FarmCPU and BLINK demonstrated superior computational efficiency, with runtimes of 10.03 ± 4.37 and 10.61 ± 3.67 seconds, respectively, compared to CMLM (52.86 ± 1.37 seconds). Manhattan and QQ plots confirmed better false positive control and p-value calibration for MLMM, FarmCPU, and BLINK. Conversely, traditional mixed models (MLM, CMLM) showed higher Type I error rates, and GLM exhibited elevated false positives. These findings underscore the robustness of multi-locus methods for wheat GWAS, particularly for traits with moderate to high heritability, and provide actionable guidelines for method selection, emphasizing MLMM for maximal power, FarmCPU for balanced performance, and BLINK for rapid, large-scale analyses. However, limitations in modeling polygenic traits and epistatic interactions, alongside wheat’s high linkage disequilibrium and polyploid nature, highlight the need for further research across diverse genetic architectures.