Implementing genomic selection in the IMB: challenges and opportunities

Stefano Biffani, Mayra Gómez Carpio, Roberta Cimmino, Dario Rossi, G. Zullo, Riccardo Negrini, Alberto Cesarani, Giuseppe Campanile, Gianluca NegliaRevista 2023 · DOI

Abstract

This study evaluates the feasibility and impact of implementing the single-step genomic best linear unbiased predictor (ssGBLUP) for genetic evaluation in the Italian Mediterranean Buffalo (IMB). By integrating genomic data with traditional pedigree information, the research demonstrates significant improvements in the accuracy of breeding value (BV) estimations for production and morphological traits. Notably, economically vital traits like mozzarella yield (MY) saw accuracy increases of up to 41%. These results suggest that ssGBLUP is a superior methodology for enhancing genetic progress in the IMB population compared to traditional pedigree-based methods.

Introduction

The Italian Mediterranean Buffalo (IMB) industry is a cornerstone of high-value dairy production, particularly for the world-renowned Mozzarella di Bufala Campana. Historically, genetic improvement within this breed has relied on traditional pedigree-based selection methods, such as the Best Linear Unbiased Predictor (BLUP). However, these methods are often hampered by incomplete genealogical records and the inherent reproductive challenges of buffalo. Genomic selection, which utilizes DNA markers (SNPs) to predict genetic merit, offers a transformative solution. This paper details the implementation of a "single-step" genomic approach (ssGBLUP) and its effectiveness in improving the accuracy of breeding value predictions for the IMB.

Methodology

The researchers compared two primary genetic evaluation models:

  1. Traditional BLUP: Utilizes the numerator relationship matrix (A) based solely on known pedigrees.
  2. ssGBLUP (Single-step Genomic BLUP): Blends the pedigree matrix (A) with a genomic relationship matrix (G) to create a unified "H" matrix. This allows the model to estimate breeding values for all animals, including those that have not been genotyped, by leveraging the genomic information of their relatives.

Validation was conducted using the Linear Regression (LR) method. A "partial" dataset (cutting off records at 2012) was used to predict performance in a "whole" dataset (records up to 2022). The models were evaluated based on correlation (stability), accuracy, dispersion, and bias.

Data and Population

The study utilized extensive records provided by the National Association of Italian Buffalo Breeders (ANASB):

  • Production Records: 743,904 lactations from 276,451 cows born between 1984 and 2019.
  • Morphological Records: 91,966 cows (2004–2022) focusing on feet, legs, and mammary systems.
  • Genotypes: DNA data from 2,017 buffalo cows and 133 bulls.
  • Traits Analyzed: 270-day milk yield, mozzarella yield (MY), fat and protein percentages/weights, and morphological "type" traits.

Results and Discussion

The transition to ssGBLUP yielded substantial quantitative improvements across all analyzed traits. For production traits, the correlation between estimated breeding values (EBVs) from partial and whole datasets increased from 6% to 49%, indicating much higher prediction stability.

Accuracy Gains

The most significant impact was observed in the accuracy of EBVs, which increased between 6% and 41% depending on the trait. Mozzarella Yield (MY), a trait of paramount economic importance, experienced the most dramatic improvement with a 40.68% increase in accuracy over traditional BLUP. Other production traits, such as protein and fat content, also saw accuracy gains exceeding 20%.

Model Stability and Bias

The ssGBLUP model demonstrated significantly less bias and better dispersion (closer to the ideal value of 1.0) than the traditional BLUP. Importantly, these improvements were also observed for non-genotyped females, proving that the unified "H" matrix effectively propagates genomic information throughout the entire pedigree.

The Role of Female Genotyping

A key insight from the study is the importance of genotyping females. In the IMB industry, where relatively few authorized bulls are used for artificial insemination, genotyping cows is essential for providing the data density required to drive genetic progress. Furthermore, the results indicate that traits with lower heritability or those that are difficult to measure traditionally benefit most from the integration of genomic data.

Conclusion

The implementation of ssGBLUP is not only feasible but highly effective for the genetic evaluation of the Italian Mediterranean Buffalo. By providing more accurate and stable breeding values, this methodology enables more precise selection decisions, ultimately enhancing the productivity and economic sustainability of the IMB industry. Future work should focus on expanding the genotyped population and integrating ssGBLUP into routine national genetic evaluations.

Summary updated

Abstract

Single-step genomic best linear unbiased predictor (ssGBLUP) is a method for jointly estimating breeding values (BV) for genotyped and non-genotyped animals. Genomic information in the Italian Mediterranean Buffalo (IMB) is now available. Its inclusion in the genetic evaluation system could increase both the accuracy and genetic progress of the traits of interest of the breed. The study aimed to test the feasibility of ssGBLUP and show the first results of implementing a genomic evaluation for production and type traits in the IMB. Phenotypic information on production (270-day milk, mozzarella yield (MY), protein and fat kg and %, respectively) and morphology: feet and legs (FL) and mammary system (MS) were used for this study. Production records included 743,904 lactations from 276,451 buffalo cows born from 1984 to 2019. Morphological traits were from 91,966 buffalo cows from 2004 to 2022. Regarding the genotypes, 2,017 buffalo cows and 133 bulls were used. Data were analyzed fitting two multi-trait animal models, a 6-trait model for production data and a 2-trait model for morphology data. According to the relationship matrix used, two models were fitted: (i) the BLUP with the numerator relationship matrix (A) and (ii) the ssGBLUP where A and the genomic relationship matrix (G) are blended into H. BVs were estimated with BLUP and ssGBLUP models. The cutoff year used to create the partial data set was 2012. The correlation, accuracy, dispersion, and bias statistics were calculated (LR method). Both bulls (N=49) and cows (N=1288) were used for validations. On average, the correlation between EBVs from partial and whole datasets estimated with BLUP and ssGBLUP increased from 6 to 49% and from 14 to 17% for production and type traits, respectively. Among the traits analyzed, the most affected by the change were protein/fat content, MY, and AM. The accuracy increase for these traits was above 20% when using the ssGBLUP. All LR statistics also improved for non-genotyped females. These results showed that implementing ssGBLUP in the breeding program can generate more accurate predictions for essential traits in dairy IMB than traditional BLUP.
BibTeX
@inproceedings{biffani2023implementing,
  author = {Stefano Biffani and Mayra Gómez Carpio and Roberta Cimmino and Dario Rossi and G. Zullo and Riccardo Negrini and Alberto Cesarani and Giuseppe Campanile and Gianluca Neglia},
  title = {{Implementing genomic selection in the IMB: challenges and opportunities}},
  booktitle = {Revista},
  year = {2023},
  doi = {10.52973/rcfcv-wbc013},
}
Figure