As can be seen in Table 3, in the genomic evaluation as the level of artificial insemination increased, the average of true breeding value (TBV) changed and genetic improvement showed a rising trend so that the highest value was obtained for the level of 100% of artificial insemination (148 kg per generation). In four studied genomic evaluation programs, the increase in the average of true genetic value from generation 0 to 5 was obtained 2, 1.2, 2.2 and 4 times, respectively. The effect of artificial insemination in the genomic selection program was evident in the distribution of superior genetic material and elevated the amount of real breeding value and the amount of milk production. Overall, the observed improvement indicates that an increase in the number of artificially inseminated females contributes to greater genetic merit for the milk production trait, regardless of the recording program. Also, using the progeny test strategy and implementing the recording system in herds can significantly enhance the amount of production (Seno et al. 2012). Figure 1 also presents the phenotypic trends of four evaluation programs over 5 generations of selection, which corresponded to the results of actual breeding value changes. However, to compare classical evaluation with genomic evaluation, two artificial insemination levels of 100% were considered. In fact, classical and genomic evaluations have been used in any of the two programs to select males, respectively. As seen in Table 3 and Fig. 1, genomic evaluation with 100% artificial insemination level showed greater genetic and phenotypic improvement. These results are consistent with those reported in other research, comparing classical and genomic evaluation (Seno et al. 2018). In the present study, genomic selection showed an increase of 1.4 times compared with the classical evaluation. In the study of dairy cows, an increase of 9% has been reported for genomic selection using a high-density panel (GBlup) and a low-density panel (GBlupi), compared to the classical evaluation (Seno et al. 2017). In general, since the evaluation and selection of buffaloes in genomic evaluation have been carried out based on their genome, the accuracy of selection was higher than that in classical selection. In fact, in this comparison, the best performance was observed for genomic selection because of genomic evaluation based on phenotypic information as well as genotypic information obtained from molecular markers that are in linkage disequilibrium with the desired genes. However, in classical evaluation, the selection is performed based only on phenotypic and pedigree information and estimated breeding values. Therefore, the performance improvement resulting from the genomic selection can be related to the better estimation of the genetic competence of the studied population. The results of the simulated population reports have shown that the correctness of the estimated genomic breeding values for male calves can be equal to the correctness of the estimated breeding values after progeny testing and through classical selection. Genomic selection could potentially lead to a doubling of genetic progress through selection and mating male buffaloes at the age of 2 years old instead of 6 or higher (Schaeffer 2006; Smaragdov 2009).
A total of 498 genotyped animals at 49,164 loci and 80,417 test day records of milk, fat and protein yields from 4127 cows were used to test the feasibility of genomic selection in the Italian Mediterranean water buffalo. Also, in this study, 50 buffaloes with phenotypes and genotypes were selected as validation animals, and variance components were estimated with BLUP and ssGBLUP. They concluded that, even if based on a small number of animals, to implement a genomic evaluation in the Italian buffaloes, female genotyping could enhance the accuracy of the prediction compared to both pedigree-based and genomic evaluations with only male genotypes models, but the inclusion of genotypes only for bulls (i.e. ssGBLUP-A) did not lead to any improvements compared to BLUP (Cesarani et al. 2021). In other studies, in carrying out the simulation analysis using different approaches which involve two types of SNP density (54 K and 100 K) and three levels of heritability traits (h2 = 0.1, 0.3, and 0.5) it has been suggested to design GS strategies by considering the SNP density and trait heritability to achieve long-term and sustainable genetic gain and to effectively control inbreeding levels (Zheng et al. 2022).
Also, since progeny testing is a costly program, breeding programs that eliminate progeny testing will be less expensive. One way to reduce the costs of a breeding program is when genomic selection is used for the initial selection of males, which will reduce the number of males required in progeny testing (Michelizzi et al. 2011). In the study of the application of genomic selection in the breeding perspective of NiliRavi buffaloes of Pakistan, it has been reported that genomic selection could potentially reduce the generation interval in the male-to-male selection pathway from 9.5 years to 3.3 years (Moaeen-ud-Din and Bilal 2016). This selection can almost double the selection response compared to the progeny test program. In addition, it will reduce the cost of male buffalo proofing by 88%. In the review of 4 breeding programs: 1) with common progeny testing, 2) absence of progeny testing and the employment of genomic selection, 3) mere genomic selection and 4) pre-selection with genomics information and then the progeny testing method in the selection of superior males, it has reported that there was a positive effect on annual genetic improvement between the use of GEBV and short generation interval. In the study of these researchers, the presence of a genomic selection plan and the absence of progeny testing had a significant effect on annual genetic improvement being among the top four plans. In addition, the inbreeding rate was less than 1% per generation (Buch et al. 2012).
Inbreeding
At first in genomic evaluation, as can be seen in Table 3, the average inbreeding coefficient has decreased from the artificial insemination level of 20% to the level of 50%. It can be said that at the artificial insemination level of 50%, half of the population was artificially inseminated using the selected best males, and for the other half of the population, mating was performed naturally. But at the level of 20%, a smaller proportion of the population was subjected to artificial insemination with 4 selected buffalo males, which enhanced the inbreeding coefficient and as a result, the average total inbreeding of the population was more than that in 50% level of artificial insemination.
In contrast, this level of inbreeding has shown a sudden increase at a level of 80% and then 100%, which is due to a greater population with artificial insemination and fewer buffalo males. However, several studies have shown that breeding programs that enhanced genetic progress usually increase inbreeding (Quinton et al. 1992; De Boer and Van Arendonk 1994). Also, in investigating genetic gain and inbreeding by genomic selection in Australian national herds it has been shown that GS may elevate the inbreeding coefficient of the offspring populations after many generations of its application (Scott et al. 2021). The maximization of the genetic gain obtained by GS is made through high selection intensity, which might reduce genetic diversity and enhancement in inbreeding ratio (Curik et al. 2014; Howard et al. 2017).
In comparing two classical and genomic evaluation programs with the same level of artificial insemination and the same number of selected buffalo males, the results showed that the average inbreeding coefficient in genomic selection was 11.3% which was less than the classical selection. Notably, Mendelian sampling was reported as the main reason for this reduction in the inbreeding rate in genomic selection. This can more effectively differentiate within families and fewer sets of half and full-sibs, thereby reducing inbreeding. According to this report, in genomic selection, the fraction of additive variance between families decreases rapidly due to the accuracy of the estimated breeding values and highlights Mendelian sampling, which does not affect the inbreeding rate (Daetwyler et al. 2007). A 10% reduction in the average inbreeding coefficient has also been reported for genomic evaluation compared with classical evaluation for dairy cows (Seno et al. 2018).
Income and cost of the studied programs
In genomic evaluation, comparing the cost of programs with different levels of artificial insemination, as can be seen in Fig. 3, the costs were higher with more artificial insemination and the highest cost was observed for the genomic selection program with 100% artificial insemination. In the studied population with genomic selection, genotyping was the most expensive item followed by artificial insemination. In this respect, with increasing the level of artificial insemination from 20 to 100% (4 programs), artificial insemination and genotyping costs for the studied population were 1.70, 2.77, 3.26 and 3.44%, and 34.57, 55.57, 65.37 and 69.03% of total selection program costs, respectively. Since with more artificial insemination, the genetic progress has also been improved for milk production, higher income from the milk yield enhancement has compensated for the effect of increasing the artificial insemination and genotyping costs. Therefore, these two costs have increased for the levels of 80 and 100% artificial insemination by a much lower amount than the two levels of 20 and 50% artificial insemination.
* CS100 = classical selection with 100% artificial insemination, GS20 = Genomic selection with 20% artificial insemination, GS50 = Genomic selection with 50% artificial insemination, GS80 = Genomic selection with 80% artificial insemination and GS100 = Genomic selection with 100% artificial insemination.
In comparing two genomic and classical evaluation programs, in genomic evaluation with 100% artificial insemination level, the genotyping cost was $1.20 million for the studied selection period, which includes 69.03% of the total costs of implementing the breeding program ($1.74 million). However, in the classical evaluation, there was no cost of genotyping. On the other hand, the cost of recording during the implementation of the program for 5 generations has been the highest (94.43%), making it more expensive than genomic evaluation. In general, despite the small number of buffalo for genotyping in the studied programs, it can be said that the rural residents will need much capital for investment. Nevertheless, when the investment in buffalo genotyping is considered a part of the investment or establishment of the dairy livestock, such investment is economically justifiable (Seno et al. 2018).
In the study of the genomic selection for Nili Ravi buffaloes in Pakistan, it has been reported that male buffalo proofing costs were reduced by 88% with this evaluation method, and initiation of a genomic program for Pakistani buffaloes has been suggested (Moaeen-ud-Din and Bilal 2016).
Regarding the total income and revenue sources, the amount of revenue from milk production during the 5 studied generations was 15.8 and 19.31 percent of the total income for the two classical and genomic evaluation programs, respectively. As can be seen in Fig. 3, the income obtained from the genomic evaluation was higher than the classical evaluation, which was due to greater genetic progress for this evaluation method. In addition, the artificial insemination program in buffalo is not as widespread as in cows (Fig. 3). Therefore, despite the greater income from the sale of milk for genomic evaluation, in general, the income was not significantly different between the two classical and genomic evaluation programs and the income from classical selection belonged to the two methods of 100% artificial insemination and 80% artificial insemination in genomic selection.