Bioscience Evidence 2025, Vol.15, No.5, 249-259 http://bioscipublisher.com/index.php/be 251 The commonly used specific methods at present include support vector machines, random forests, gradient boosting and deep neural networks, etc. These algorithms and deep learning (DL) models have been widely applied in various techniques such as genotype-phenotype prediction, gene mining, and multitrait improvement (Galli et al., 2021; Yan et al., 2021; Kudiyarasudevi and Suresh, 2024; Wu et al., 2025). Platforms similar to AutoGP combine multiple ML and DL models, allowing users to select the most appropriate algorithm for faster and more accurate genomic selection. meta-ensemble learning integrates different algorithms to make multi-trait prediction more stable and better adaptable to different situations Figure 1 Relationship between AI, ML, and DL (Adopted from Cravero et al., 2022) 3.2 Genomic selection and predictive breeding Genomic selection (GS) uses whole-genome markers and statistical or machine learning models to predict an individual's breeding values. This can shorten the breeding time and also improve genetic progress (Tong and Nikoloski, 2020; Xu et al., 2022; Mora-Poblete et al., 2023; Barreto et al., 2024; Kudiyarasudevi and Suresh, 2024; Wu et al., 2024; He et al., 2025; Wu et al., 2025). Nowadays, some new methods, such as deep learning, automated machine learning (AutoML), and ensemble learning, can handle high-dimensional and multi-type data. These methods have significantly improved the prediction accuracy of complex traits such as yield and stress resistance. For example, after the AutoML framework combines the dimensionally reduced environmental parameters and feature labels, the prediction accuracy can be increased by 14% to 28%, which is helpful for the development of climate-adaptive corn varieties (He et al., 2025). Meanwhile, deep learning models with multiple traits and multiple environments perform better than traditional Bayesian models and linear hybrid models in the prediction of complex traits such as flowering period (Mora-Poblete et al., 2023). 3.3 Data visualization and decision-support systems With the increasing volume and complexity of data, the role of data visualization and decision support systems (DSS) in corn breeding is becoming more and more important. Modern intelligent breeding platforms (such as AutoGP, CropGBM, etc.) all come with visual interfaces. They can present multi-dimensional data such as genotype, phenotype and environment in an interactive way, helping researchers intuitively understand the prediction results and key factors (Yan et al., 2021; Wu et al., 2025). These systems not only improve the efficiency of data interpretation, but also provide references for breeding decisions, such as selecting parental combinations and screening superior strains. In addition, the decision support system driven by big data can also automatically recommend the best breeding plan, promoting intelligent and precise breeding (Esposito et al., 2019).
RkJQdWJsaXNoZXIy MjQ4ODYzNA==