Comparison of breeding strategies for the creation of a synthetic pig line

Creating a new synthetic line by crossbreeding means complementary traits from pure breeds can be combined in the new population. Although diversity is generated during the crossbreeding stage, in this study, we analyze diversity management before selection starts. Using genomic and phenotypic data from animals belonging to the first generation (G0) of a new line, different simulations were run to evaluate diversity management during the first generations of a new line and to test the effects of starting selection at two alternative times, G3 and G4. Genetic diversity was characterized by allele frequency, inbreeding coefficients based on genomic and pedigree data, and expected heterozygosity. Breeding values were extracted at each generation to evaluate differences in starting selection at G3 or G4. All simulations were run for ten generations. A scenario with genomic data to manage diversity during the first generations of a new line was compared with a random and a selection scenario. As expected, loss of diversity was higher in the selection scenario, while the scenario with diversity control preserved diversity. We also combined the diversity management strategy with different selection scenarios involving different degrees of diversity control. Our simulation results show that a diversity management strategy combining genomic data with selection starting at G4 and a moderate degree of diversity control generates genetic progress and preserves diversity.


50
A specific breeding scheme can be established after two or three generations of mixing 51 (Legault et al., 1996). However, the original genetic diversity has to be preserved in order 52 to promote the complementarity of the genomes of the original breeds and to limit genetic drift 53 towards one of the parent breeds, particularly through selection (Paim et al., 2020). Knowledge 54 of the genomic composition of crossbred animals would thus help keep the genomic composition 55 between the parental breeds balanced over generations. The mating plan for breeding stock is 56 also an important step to favor haplotype recombinations and limit inbreeding. Inbreeding can 57 be characterized at the genomic level by different methods. One of them is the detection of 58 runs of homozygosity (ROH), continuous homozygous chromosomal segments along the genome Figure 1: Simulation scenarios. In step 1 we compared our diversity management strategy hereafter named 'Genomic strategy', with a 'Random scenario' and a 'Selection on BV' scenario. In step 2, we tested two selection scenarios with two selection starting points, i.e., G3 and G4. Three different scenarios were computed to compare a strategy using genomic data to max-113 imize diversity with two alternative scenarios. The first scenario is a basic simulation with both reproducers and mating selected randomly, 116 called Random scenario. We repeated this scenario 25 times.  (Meuwissen 2001) in MoBPS with known heri-121 tabilities. We selected reproducers with the lowest BV because the objective of selection was 122 to improve the growth rate by minimizing A100. We computed a coancestry coefficient f P ED 123 based on pedigree data with function pedIBD() in the optiSel R package (Wellmann, 2019). 124 This coefficient was based on the pedigree of real G0 animals (over 10 generations) concate-125 nated with the simulations without errors. A mating plan was established with an optimization 126 algorithm (in-house Fortran script) based on the minimization of f P ED sum between matings.

127
This scenario represented a selection strategy that maximizes genetic gain, where genomic data 128 were used to obtain accurate BV, and mating was managed by minimizing pedigree relationship.

129
Selection on BV was repeated 25 times. The third scenario, called Genomic strategy, is based on a strategy where genomic data 132 are used both for selection and genetic diversity management. Two principles were combined to 133 select reproducers. The first was to choose reproducers among animals with balanced genomic 134 composition between founding breeds, with a particular focus on the proportion of Duroc origin.

135
The second principle was to promote original animals in terms of alleles. An animal is original 136 if it carries alleles that are uncommon in the population.

137
Genomic compositions between PLW and Duroc were calculated based on the real points 138 of recombination in each meiosis. Note that this is not possible in practice and needs to be 139 estimated and thus represents an idealized estimate compared to a real-world scenario. If a 140 chromosomal fragment originated from the maternal chromosome of a G0 animal, its origin was 141 a PLW founder. Otherwise, if a chromosomal fragment originated from a paternal chromosome 142 of a G0 animal, its origin was a Duroc founder. We considered an animal to be balanced when animal i, ORI index was computed with in-house Python script with the following formula: computed as: where L ROH is the sum of the length of all the ROH detected in an animal in bp, and L autosomes 207 is the total length of the autosomes covered by markers in bp.     The two graphs with genomic coefficients F ROH and F M were similar ( Figure 3A,B). They 244 showed a significant increase in the two coefficients in the Selection on BV scenario. Inbreeding 245 increased in the Random scenario but to a lesser extent than in the Selection on BV scenario 246 and its increase was strictly linear from G1 to G10. The most limited increase in inbreeding was 247 observed in the Genomic strategy scenario. F ROH increased slightly between G0 and G1 and 248 decreased slightly after G1 and then increased from G3 to G10. F M increased between G0 and 249 G1 like F ROH but after G1, it decreased until G7 and then appeared to stabilize. In G10, F ROH and Genomic strategy scenarios, respectively.

253
Unlike F ROH or F M , F P ED increased significantly in all three scenarios Figure 3C, albeit 254 with a slightly different pattern. In the Selection on BV scenario: between G0 and G3, F P ED 255 remained the lowest, with a bigger increase after G3 to reach the highest F P ED value (0.06) in 256 G10. In the Random scenario, F P ED presented a strictly linear increase after G1. The profile 257 of the Genomic strategy was similar to that in the Selection on BV between G0 and G3, 258 subsequently, F P ED increased at a slower rate and at G10, this scenario had the lowest F P ED

Expected heterozygosity
In all three scenarios, we observed an increase in He at G1 that was higher in the Genomic 263 strategy and Selection on BV scenarios (Figure 4). In the following generations, He de- Here we compared three selection scenarios using two different generations to start the 272 selection, G3 and G4. Genetic diversity estimators and BV were analyzed from G3 to G10 or 273 G4 to G10, respectively.  Differences between starting selection in G3 or G4 were negligible in the two scenarios. The 281 shape of the two respective curves for the allele frequency contributions was similar at both 282 starting points, with the G3 curve basically shifted by one generation (to the left) relative to 283 the G4 curves.

284
The third scenario BV and pedigree data, showed the strongest decrease in the proportion 285 of Common alleles compared to the two other scenarios with genomic data and ORI index. The 286 curves for the start of selection at G3 and G4 were again similar, however, as the shift per 287 generation in BV and pedigree data was highest the biggest differences in G10 were observed.

288
There was a slight increase in Intermediate and Rare alleles that was similar in the two starting 289 points.

290
The proportion of Fixed alleles, was constant in the three scenarios.

292
The evolution of inbreeding was characterized with the simulation of selection scenarios. As 293 mentioned above, we studied F ROH , F M and F P ED (Figure 6). Figure 6A shows the results for 294 F ROH were similar in the two scenarios with genomic data and ORI index. F ROH first increased, 295 then decreased, and then increased again. The shape of the curve was the same whether selection 296 started at G3 or G4, the two curves only shifted by one generation. However, at G10, in the 297 two scenarios, the simulation that started at G3 had a higher F ROH than the simulation that 298 started at G4. Regarding the difference between the scenarios BV + genomic data with 299 relaxed ORI and BV + genomic data with stringent ORI, in limiting inbreeding, the 300 second scenario was more efficient. In the third selection scenario, BV + pedigree data, 301 F ROH increased immediately after the start of selection until G10. The increase was lower when 302 selection started in G4 than when it started in G3.

303
The second genomic-based inbreeding coefficient was F M (Figure 6B). The results for this 304 coefficient were similar to previous results for F ROH . Once again, starting selection at G4 limited 305 the increase in inbreeding better. 306 Last, we studied F P ED (Figure 6C), the shapes of the curves in the scenarios with genomic 307 data with ORI index were similar. The results of the scenario BV + pedigree were very close 308 results for the two selection starting points.

309
For the three coefficients, the scenario BV + genomic data with stringent ORI limited 310 inbreeding the most. However, in the first generations of selection (G4 to G6), the scenario BV 311 + pedigree with a start of selection at G4 showed the lowest level of inbreeding. The estimated genetic progress was compared in the three selection scenarios (Figure 7).

314
The selection applied in the three scenarios aimed to decrease the BV of the studied trait (A100).

315
In all scenarios and in line with the breeding objective, BV decreased over the 6 or 7 simulated 316 generations. These results highlighted improved growth of simulated animals. In the BV + 317 genomic data with relaxed ORI and BV + pedigree data scenarios, there was a shift of 318 approximately one generation between starting selection at G3 and at G4 in BV evolution. This 319 result was not observed for BV + genomic data with stringent ORI where the improvement 320 of BV from G7 was somewhat limited when selection started at G4.

321
As expected, in G10, the best improvement of BV was observed for BV + pedigree data,  In the three scenarios in which we tested two selection starting points, our results suggest 498 that, in the context of diversity analysis, it would be advantageous to start selection at G4 499 rather than G3. One additional generation managed with Genomic strategy allowed further 500 chromosomal recombinations that increase diversity within the population. On the other hand, 501 starting selection at G3 is more efficient and generates genetic progress more rapidly than starting 502 at G4. Not surprisingly, the higher the pressure on diversity, the lower the pressure on genetic 503 progress. Therefore, the first three generations of a new line managed with Genomic strategy 504 followed by BV + genomic data with relaxed ORI appears to be a good compromise 505 between conserving diversity and achieving genetic progress. 506 We did not analyze a scenario with OCS in this paper. This method of selection uses the av- Genomic data offer new opportunities for genetic diversity management in composite breeds.

517
Our results show that diversity management based on genomic data can be used in the first 518 generations of a new line to build diversity even before starting selection. In the case of two 519 step three-way crossbreeding, it was clearly better to start the selection after 4 generations 520 post-crossbreeding to limit inbreeding, however it is difficult to extend this recommendation to 521 all kinds of crossbreeding strategies. In any case, simulations based on real data collected from 522 the base crossbred population should be carried out to design an appropriate breeding strategy 523 that combines selection and diversity management. Among the breeding scenarios studied here, 524 a strategy that includes a focus on allele frequencies, such as the BV + genomic data with 525 relaxed ORI would be a good compromise between genetic progress and diversity conservation. We thank all members of NUCLEUS R&D and technical services as well as the breeders 533 involved in this project. We thank Jean-Michel Elsen for his advice.