Introduction This study proposes a deep learning based framework to investigate genomic factors influencing athletic performance in sports education. Traditional approaches often face challenges in modeling the high dimensionality and complex genotype phenotype relationships inherent in genomic data. Methods To address these issues, the proposed framework integrates three core components: a formalized problem formulation, a Genomic Athletic Predictor, and a Constrained Optimization Refinement mechanism with uncertainty aware prediction. The method models the genomic feature space and athletic performance metrics under manifold informed constraints, while explicitly incorporating uncertainty quantification to enhance reliability. The Genomic Athletic Predictor consists of a Manifold Informed Constraint Encoder, an Agent Driven Genomic Planner, and an Uncertainty Guided Athletic Forecaster, enabling structured representation learning and robust performance prediction. Results and Discussion Experimental evaluations conducted on two large scale cohort datasets demonstrate that the proposed framework consistently outperforms classical regression models, ensemble learning methods, and advanced neural network baselines. The model achieves superior results across multiple evaluation metrics, including Pearson correlation, RMSE, R 2 , and MAE, while maintaining moderate computational complexity. Ablation studies further confirm the complementary contributions of manifold constraints and uncertainty modeling in improving predictive stability and biological plausibility. These findings highlight the effectiveness of integrating deep learning, domain constraints, and uncertainty aware modeling for genomic based athletic performance prediction, offering practical implications for personalized training, talent identification, and data driven optimization in sports education.
Dong et al. (Thu,) studied this question.