Mean Normalization Improved Genome-Wide Association Detection Power of Wheat (Triticum aestivum) Grain and Flour Quality Traits with Year-to-Year Variation

Abstract

Listen

Translate article

Grain and flour quality traits affect marketing potential and milling and baking properties. Trait means varied in fourteen wheat grain and flour quality traits for a population of 188 diverse soft winter wheat varieties harvested from 2020 to 2023 at the same location. Significant weather differences occurred yearly. This created a challenge for the detection of chromosome locations affecting these traits through genome-wide association studies (GWAS). Mean normalization using standard deviation to transform raw data to Z scores has been used successfully in other statistical analyses of biological systems with mean differences. Mean normalization was applied to a GWAS, improving detection power for thirteen grain and flour quality traits with high broad-sense heritability. It did not improve the lone trait with low heritability. Improvement was measured as the reduction in the p-value of mean normalized data compared with raw data for the same significant marker using the same GWAS model in the same trait. Improvement varied by trait and marker, but the average p-value of 135 common significant marker/GWAS model combinations was reduced 27 times with mean normalization over raw averaged data. Mean normalization reduced p-values ~1800 times when compared with a GWAS using best linear unbiased predictors. However, the best linear unbiased predictors led to only 15 common marker/GWAS model combinations with mean normalization, limiting the ability for direct marker comparison. Test weight, kernel protein, kernel weight, sodium carbonate solvent retention capacity, and sucrose solvent retention capacity showed the greatest increased detection power.