Pipelines for data analysis for manuscript Missing Comparability: When genomic selection faces field variability. A case study in soybeans.

Genomic selection is a method to predict how crops will perform based on their DNA. This process is often affected by environmental factors, such as soil quality and weather, which can mask a plant's true genetic potential. In this study, we tested several methods to separate genetic signals from field variability using real soybean data and computer simulations. Our results show that cleaning the data reduces models’ ability to predict phenotypes, but simulations prove that it actually makes them much more accurate approximating true genetic values. We conclude that current genomic models are more accurate than they appear, and their predictions can be a better selection criterion than the observed phenotypes.