Genomics and Applied Biology 2024, Vol.15, No.6, 307-319 http://bioscipublisher.com/index.php/gab 311 The creation of comprehensive marker databases, such as the Yam Microsatellite Markers Database (Y2MD), has further streamlined the breeding process. Y2MD provides breeders with rapid access to an extensive collection of SSR markers, promoting data sharing and collaboration within the yam breeding community. This shared resource allows for more efficient screening and selection across breeding programs, fostering the development of resilient yam cultivars suited to diverse environmental conditions and agricultural challenges (Diouf et al., 2023). As these resources continue to grow, they play a pivotal role in enhancing the precision and speed of yam breeding efforts globally. 4.3 Role of genomic resources in yam breeding Genomic resources have transformed yam breeding by offering critical tools for identifying genetic diversity and advancing the development of resilient yam cultivars. These resources, which include comprehensive databases of genetic markers, allow breeders to employ both marker-assisted selection (MAS) and genomic selection (GS) with high precision. By integrating MAS and GS, breeders can significantly streamline the selection process, allowing for faster breeding cycles and more targeted trait selection. This approach is particularly valuable for enhancing traits like adaptability, yield, and disease resistance, which are essential for ensuring sustainable and productive yam varieties under varying environmental conditions. The availability of genotyping technologies, such as genotyping-by-sequencing (GBS), simple sequence repeats (SSRs), and single nucleotide polymorphisms (SNPs), has greatly supported trait-specific breeding efforts. With these tools, breeders can make data-driven decisions, especially for complex, polygenic traits like yield and disease resistance, which are typically challenging to select using traditional methods alone. By enabling precise trait mapping and early prediction of genetic potential, these genomic resources ensure that yam breeding programs are better equipped to meet agricultural demands, promoting food security and sustainability in yam-producing regions (Mulualem et al., 2018). 5 Development and Optimization of Genomic Prediction Models The development and optimization of genomic prediction models are fundamental for the successful application of genomic selection in yam breeding. These models enable breeders to predict the genetic potential of breeding candidates based on their genomic profiles, allowing for earlier and more precise selection. Genomic prediction models can be tuned and improved to better predict complex traits, which are common in yam, including yield, disease resistance, and adaptability. Each model has its own strengths and limitations, and careful selection of the right model for specific traits and conditions is crucial to maximize genetic gains and reduce breeding cycle time. 5.1 Commonly used genomic prediction models and their principles Genomic prediction models are essential tools in plant and animal breeding, allowing the estimation of breeding values based on genome-wide markers. These models vary significantly in their assumptions and mathematical approaches. One of the most widely used models is the best linear unbiased prediction (BLUP) approach, particularly genomic BLUP (GBLUP). GBLUP estimates breeding values by incorporating a genomic relationship matrix derived from SNP (single nucleotide polymorphism) data. This approach assumes that genetic effects are additive and normally distributed. GBLUP is efficient for high-dimensional data as it treats all markers as having small, equal effects, which is useful for predicting polygenic traits (Ge et al., 2020). Bayesian models are another common class of genomic prediction models, including methods like Bayes A, Bayes B, and Bayes Cπ. These models differ from GBLUP in that they allow some markers to have larger effects than others, which can be advantageous for traits controlled by a few major genes. Bayes A assumes that all markers have a similar prior variance, while Bayes B and Bayes Cπ use a mixture prior where some markers are assumed to have no effect. This flexibility enables Bayesian models to capture genetic architectures where only a subset of markers contributes significantly to trait variability (Crossa et al., 2017). In addition to these traditional models, nonparametric and machine learning-based approaches are increasingly applied in genomic prediction. Support vector machines (SVM) and deep learning models, like convolutional neural networks (CNN) and multilayer perceptrons (MLP), are examples. These methods do not assume an
RkJQdWJsaXNoZXIy MjQ4ODYzMg==