Computational Molecular Biology 2024, Vol.14, No.2, 54-63 http://bioscipublisher.com/index.php/cmb 58 effects by treating them as random variables with specific prior distributions. This approach enables the modeling of complex genetic architectures and the incorporation of uncertainty in parameter estimates. Bayesian methods have shown superior performance in predicting breeding values, particularly for traits with low heritability, by effectively capturing the underlying genetic variance (Muir, 2007; Zhang et al., 2010). One of the key advantages of Bayesian methods is their ability to handle large-scale genomic data and provide more accurate predictions compared to traditional methods. For instance, the BayesB method, which assigns a mixture of distributions to marker effects, has been shown to outperform other BLUP-based methods in terms of prediction accuracy. This is particularly evident in scenarios where the genetic architecture of the trait involves a few large-effect loci and many small-effect loci. The flexibility of Bayesian methods in accommodating different genetic architectures makes them a powerful tool for genomic selection in both plant and animal breeding (Muir, 2007; Zhang et al., 2010). 4.3 Machine learning algorithms Machine learning algorithms have emerged as powerful tools for genomic prediction and selection, offering the ability to model complex, non-linear relationships between genotypes and phenotypes. Techniques such as random forests, support vector machines, and neural networks have been applied to genomic data to improve the accuracy of breeding value predictions. These algorithms can handle high-dimensional data and capture interactions among markers, making them suitable for predicting complex traits influenced by multiple genetic and environmental factors (Muir, 2007; Zhang et al., 2010) The application of machine learning in genomic selection has shown promising results, particularly in enhancing prediction accuracy and selection response. For example, genomic best linear unbiased prediction (G-BLUP), a ridge-regression type method, has been effectively combined with machine learning techniques to improve the prediction of complex human traits. Studies have demonstrated that machine learning algorithms can outperform traditional BLUP methods, especially when dealing with large datasets and traits with low heritability. The integration of machine learning into genomic selection frameworks holds great potential for advancing breeding programs and achieving higher genetic gains (Muir, 2007; Zhang et al., 2010; Campos et al., 2013). 5 Applications in Plant Breeding 5.1 Genomic selection for crop improvement Genomic selection (GS) has revolutionized crop improvement by enabling the prediction of breeding values using genome-wide markers. This method leverages high-density marker scores to predict the genetic potential of untested populations, thus accelerating the breeding cycle and enhancing genetic gains (Jannink et al., 2010; Desta and Ortiz, 2014; Varshney et al., 2017). Unlike traditional marker-assisted selection, which focuses on individual loci, GS incorporates all marker data, capturing the effects of small quantitative trait loci (QTL) and providing more accurate predictions (Desta and Ortiz, 2014; Varshney et al., 2017). Studies have shown that GS can achieve high correlation levels between true breeding values and genomic estimated breeding values, even for traits with low heritability, making it a powerful tool for selecting agronomic performance traits (Varshney et al., 2017). The integration of GS with advanced technologies such as high-throughput genotyping and phenotyping further enhances its efficiency and application in varietal development programs (Krishnappa et al., 2021). 5.2 Enhancing disease resistance The application of GS in enhancing disease resistance in crops has shown significant promise. By using genome-wide markers, GS can predict the genetic potential for disease resistance traits more accurately than traditional methods (Crossa et al., 2011; Wang et al., 2018). For instance, in maize, GS has been used to improve resistance to diseases such as Exserohilum turcicum and Cercospora zeae-maydis, demonstrating the method's effectiveness in real-world breeding programs (Crossa et al., 2011). The ability of GS to account for genotype × environment interactions further enhances its utility in breeding for disease resistance, as it allows for the selection of genotypes that perform well across different environmental conditions (Crossa et al., 2011; Crossa et al., 2017). This holistic approach to selection ensures that disease-resistant traits are effectively incorporated into new crop varieties, contributing to sustainable agricultural practices.
RkJQdWJsaXNoZXIy MjQ4ODYzNA==