Titel: Enhancing Predictive Accuracy of Agronomic and Quality Traits in Barley (Hordeum vulgare L.) Using Spectral Variable Selection Methods
Abstract:
Phenomic selection (PS) offers a cost-effective and breeder-friendly approach for public breeding programs with limited access to genotyping or constrained financial resources for laboratory infrastructure. As PS relies primarily on high-throughput phenotyping data, most often derived from near-infrared spectroscopy (NIRS) of seeds, prediction is challenged by the high dimensionality and complex nature of spectral data, where only a subset of spectral variables is informative and usable in models. To enhance predictability, three regularized regression models (Lasso, Elastic Net, and Ridge) were evaluated for their predictive ability on ten morphological, agronomic, and quality traits measured in two malt barley trials conducted during the 2022 and 2024 cropping seasons in Ethiopia. The models were tested across four practical breeding scenarios: within-location unseen genotype prediction(WL-uG), leave one location out prediction (LOLO), target environment unseen genotype prediction(TargetEnv), and across-location wide adaptability(RuG). Cross-validation was used to assess model performance and to select stable informative spectral predictors for each trait, scenario, and trial. Among the tested models, Lasso and Elastic Net consistently outperformed Ridge regression, with Elastic Net showing superior performance across scenarios. Prediction ability reached up to 0.83 for quality traits in TagetEnv and 0.56 for Agronomic traits. Across scenarios, RuG achieved the highest overall prediction ability, followed by WL-Ug, underscoring the importance of spectral predictor selection and transferability of PS in barley breeding:
Biography:
Tigist Shiferaw Tadesse is a doctoral candidate at the State Plant Breeding Institute of the University of Hohenheim. She holds her BSc and MSc in plant science and plant breeding, and she has worked as a barley breeder at the Ethiopian Institute of Agricultural Research (EIAR)Tigist Shiferaw Tadesse is a doctoral candidate at the State Plant Breeding Institute of the University of Hohenheim. She holds her BSc and MSc in plant science and plant breeding, and she has worked as a barley breeder at the Ethiopian Institute of Agricultural Research (EIAR)
