Powdery mildew resistance prediction in Barley (Hordeum Vulgare L) with emphasis on machine learning approaches.

By employing machine-learning models, this study utilizes agronomical and molecular features to predict powdery mildew disease resistance in Barley (Hordeum Vulgare L). A 130-line F8-F9 barley population caused Badia and Kavir to grow at the Gonbad Kavous University Research Farm on three planting dates (19 November, 19 January, and 19 March), with three replicates in 2018/2019 and 2019/2020. The study employed RReliefF, MRMR, and F-Test feature selection algorithms to identify essential phenotype traits and molecular markers. Subsequently, Decision Tree, Random Forest, Neural Network, and Gaussian Process Regression models were compared using MAE, RMSE, and R2 metrics. The Bayesian algorithm was utilized to optimize the parameters of the machine-learning models. The results indicated that the Neural Network model accurately predicted powdery mildew disease resistance in barley lines. The evaluation based on high R2 values, as well as low MAE and RMSE, highlighted the efficacy of these models in identifying significant phenotype traits and molecular markers associated with disease resistance. The findings demonstrate machine learning models' potential in accurately predicting powdery mildew disease resistance in Barley. The neural network model specifically showed excellent results in this area because it managed to identify critical phenotypic traits and molecular markers very well. This research highlights the importance of combining AI with molecular markers for improved disease resistance and other desirable crop traits during plant breeding.