International Journal of Molecular Veterinary Research, 2025, Vol.15, No.1, 13-21 http://animalscipublisher.com/index.php/ijmvr 15 2.3 Genetic differences among chicken breeds/populations and their implications for disease resistance breeding Genetic variation in disease resistance exists to a great extent between lines and breeds of chickens. The local and indigenous breeds are more resistant to certain diseases compared to commercial lines, most likely due to their greater genetic diversity and adaptation (Zhang et al., 2024). For example, Beijing You chickens are more resistant to Salmonella Pullorum than commercial Rhode Island Red strains (Li et al., 2018), and Chengkou mountain chickens possess certain avian leukosis resistance loci. These differences highlight the necessity to preserve genetic diversity and utilize resistant breeds in breeding programs in order to make the flocks more resilient (Gul et al., 2022; Li et al., 2025b). 3 Principles and Technical Methods of Genomic Selection in Chickens 3.1 Basic principles of genomic selection Genomic selection (GS) is a breeding technique that makes use of genome-wide genetic markers, typically single nucleotide polymorphisms (SNPs), to estimate the genetic value of individuals for complex traits such as disease resistance. GS is based on the application of thousands of markers spread across the genome, rather than a limited number of large genes as in the traditional marker-assisted selection, thus covering major and minor genetic effects. This allows more accurate prediction of genomic estimated breeding values (GEBVs) and accelerates genetic gain through earlier and more accurate selection of breeding candidates (Misztal et al., 2020; Wang and Wang, 2024). It is a process founded on statistical models trained on reference populations with both genotypic and phenotypic data such that the genetic architecture of the trait is well represented (Misztal et al., 2020). 3.2 Construction of reference populations and training datasets A sufficient reference population is required for effective genomic selection. The population has to be large and genetically diverse with extensive phenotypic records of target traits (e.g., disease resistance, growth, or egg production) and dense genotypic data (e.g., SNP profiles) (Wolc et al., 2016). GS accuracy is determined by the size and quality of the reference population and the accuracy of phenotypic measurement. There needs to be regular updating and retraining of prediction models with novel information to maintain prediction accuracy from generation to generation since patterns of linkage disequilibrium and frequencies of alleles may change over time (Wolc et al., 2016). 3.3 Common genotyping technologies Various genotyping technologies are used for genomic selection in chicken. The high-density SNP chip technology, such as the 50K or 600K arrays, are widely applied because of their reliability and ability to pick up the genome-wide variation, making them suitable for large-scale selection and genome-wide association studies (Wolc et al., 2016). Genotyping-by-sequencing (GBS) and restriction site-associated DNA sequencing (RADseq) are cost-effective, high-throughput next-generation sequencing approaches well suited to large or genetically diverse populations that enable effective SNP discovery and genotyping (Pértille et al., 2016; Herry et al., 2023). Whole genome sequencing (WGS) provides the highest variant information, including rare and novel SNPs, yet is more expensive and data-hungry and therefore mainly used for research and developing novel genotyping panels (Tan et al., 2023). 3.4 Statistical models and prediction methods There are various statistical models that have been used to predict genomic breeding values. The most widely used is the genomic best linear unbiased prediction (GBLUP) model, which utilizes genomic relationships among individuals to provide precise and computationally appealing predictions for intricate traits (Misztal et al., 2020; Herry et al., 2023). The single-step GBLUP (ssGBLUP) model integrates pedigree and genomic data, improving prediction accuracy, especially when animals are not genotyped. Bayesian approaches (e.g., BayesA, BayesB) allow for variable selection and shrinkage, which are suitable for traits governed by main and minor genes (Misztal et al., 2020). Machine learning techniques such as random forests and neural networks are being researched for their ability to detect complex, non-linear patterns within genomic data, while their application in poultry breeding on a routine basis is yet to surge and predominantly confined to research settings (Misztal et al., 2020).
RkJQdWJsaXNoZXIy MjQ4ODYzNA==