Improving Wheat Grain Yield Genomic Prediction Accuracy Using Historical Data.

Genomic selection (GS) is an essential tool to improve genetic gain in wheat breeding. This study aimed to enhance prediction accuracy (PA) for grain yield (GY) across various selection environments (SEs) using CIMMYT's (International Maize and Wheat Improvement Center) historical dataset. Ten years of GY data from six SEs were analyzed, with the populations of five years (2018-2023) as the validation population (VP) and earlier years (back to 2013-2014) as the training population (TP). Generally, we observed that as the number of training years increased, the PA tended to improve or stabilize. For instance, in the late heat stress SE (BLHT), PA increased from 0.11 (one training year) to 0.23 (five years), stabilizing at 0.26. Similar trends were observed in the intermediate drought SE (B2IR), with PA rising from 0.12 (one year) to 0.21 (four years) but minimal improvement beyond that. Conversely, some selection environments, such as F5IR (flat optimal environment), did not significantly increase, with the PA fluctuating around 0.09-0.14 regardless of the number of training years used. Additionally, average genetic diversity within the TP and the VP influenced PA. Indeed, a negative correlation between PA and the genetic distance was observed. This highlights the need to balance genetic diversity in TPs to enhance the predictive power of GS models. These findings exhibit the benefits of using an extended historical dataset while considering genetic diversity to maximize PA in GS strategies for wheat breeding, ultimately supporting the development of high-yielding varieties.