Genetic Improvement at the Commercial Level

Compared to Genetic Progress at the Nucleus Level

 

Jan W.M. Merks

 IPG, Institute for Pig Genetics B.V. (a TOPIGS company), The Netherlands

 

Introduction

A pig-breeding program generally consists of different levels in a pyramidal structure, indicated as nucleus, multiplication and commercial level. Selection takes place at all levels, but improvements generated in the nucleus eventually determine the rate of annual genetic change. This genetic change is economically important at all levels, but especially at the commercial level because of the relatively large number of animals.

Generally, selection at the nucleus level for growth and carcass traits is based on performance testing (both for production and reproduction traits), supplemented with family information. The testing usually takes place at the nucleus farms, sometimes central test stations, under more or less standardized environmental conditions to allow a fair comparison of the tested pigs. However, these sophisticated conditions deviate from the conditions at the multiplication level and certainly also from the conditions at commercial fattening where the breeding goal is defined. As a consequence, changes in rank order for genotypes between these environments may occur and lower the efficiency of a pig breeding program in terms of genetic progress at the commercial level.

Changes in rank order of genotypes between environments are indicated as genotype x environment interaction (G x E). Falconer (1952) proposed to measure changes in rank order between environments as the genetic correlation between phenotypes for the same genotype in different environments. This concept is based on the assumption that the expression of identical traits may in fact not be controlled by the same sets of genes if G x E exists. If the presence of G x E is just a matter of scale, thus without affecting the ranking of breeding animals, the genetic correlation equals one. In that case G x E does not affect the efficiency of the breeding program.

During the last several decades several non-unit estimates of genetic correlations between the different levels of pig breeding programs were reported. These results must be considered as serious indications of lower genetic progress at commercial level than at nucleus level for both production and reproduction traits. This paper summarizes several analyses dedicated to find the leaks in the flow of genetic progress from nucleus herds to commercial level and presents opportunities to overcome these problems.

 

Description of genotype x environment interaction in pig breeding programs

Modern pig-breeding programs are 3 or 4 breed/line crossbreeding programs. When describing the problem of genotype x environment interaction in pig breeding programs, genotype is always represented by sires, but for the environment a distinction has to be made between three categories:

 

1. Specified factors within farms such as feeding regime, housing system and sex,
2. Husbandry circumstances in general, e.g. farms, health level, level in the production pyramid, or outside climate,
3. Pure breeding (nucleus) versus crossbreeding (commercial level).

 

Any of these categories may be a factor in G x E and consequently reduce genetic progress on the commercial level. Interaction of genotype with a specific environmental factor may also be responsible for an interaction of genotype and herds if differences for this factor exist between herds. Interaction of genotype and herds may in turn be responsible for interaction of genotype and level of the breeding program, as the commercial environment includes different herds. Crossbreeding versus pure breeding is a factor by itself, but may also interact with the other factors.

Research on interaction of genotype and specific environmental factors helps to decide whether it is necessary to match these factors in the nucleus with those under commercial conditions. But not all differences in the environment between nucleus herds and commercial herds can be specified. It is even more difficult to specify differences in the environment between commercial herds.

A short description of G x E in pig breeding programs was given by Brascamp et al. (1985). On the basis of that general description, G x E can be described as in Figure 1. The three blocks represent the three levels of the breeding program; nucleus herds with performance testing, multiplier herds with on-farm testing and commercial herds with fattening pigs. The genotype x environment interaction between the different levels is represented by r_G, while r_g (analogous to the description of Brascamp et al., 1985) represents the genotype x environment interactions within a level of the breeding program.

Figure 1. Schematic description of the genotype x environment interaction problem in pig breeding programs (Brascamp et al., 1985)

 

Summary of different G x E studies

During the last decade several studies on the different aspects of G x E have been performed. In this paper most of these studies are summarized including the data used to enable a better understanding of the problem and potential solutions.

1. Purebred versus crossbred performance in nucleus herds (Merks and Hanenberg, 1998)

In most commercial pig breeding programs, the selection in the nucleus is based on pure line performance (PLS) to improve performance within lines/breeds. However, in commercial pork production, performance of crossbred pigs is dominantly important.  A combined crossbred and pure line selection (CCPS) method shows to be superior to PLS in many different situations (Bijma and Van Arendonk, 1998).

In order to apply CCPS successively, the genetic correlation between purebred and crossbred performance (r_pc) and the crossbred heritability (h²_c) are crucial factors (Van der Werf et al., 1994).

Merks and Hanenberg (1998) presented a) the results of the estimation of r_pc and (h²_c in the TOPIGS  breeding program for Great Yorkshire (GY-s) line performance compared to the performance of Duroc x GY-s and Pietrain x GY-s crossbreds and b) the consequences of these estimates for optimal selection strategy. To choose the optimal selection strategy, the GY-s breed is selected for both production of purebred GY-s as well as for crossbred terminal boars while Duroc is selected for production of crossbred boars only.

 

Data used

      Production data were used from pure line and crossbred populations of the TOPIGS sire lines Great Yorkshire (GY-s), Duroc (D), Pietrain (P) and their crosses P*GY-s and D*GY-s. Data were collected for the traits weight at 180 days (kg) and backfat thickness at 110 kg (mm). For crossbred populations, only information of traits measured on boars was available. Data from 72,419 animals, recorded from June, 1992 to May, 1997 were used in the analysis (P*GY-s data since March 1993). Data were obtained from 1 D-nucleus farm and 50 GY-s nucleus farms. Most GY-s nucleus farms were also D*GY-s or P*GY-s multiplier. Pure line information of Pietrain was not used because of their low numbers.

 

Results

Estimates of heritability for weight under a univariate model varied from 0.20 for GY-s to 0.45 for D*GYs. Estimates of heritability for backfat were higher and varied from 0.37 for GY-s to 0.62 for D. Genetic correlations between purebred and crossbred populations are given in Table 1. Genetic correlations were highest for weight and did not significantly differ from 1. Genetic correlations for backfat thickness varied from 0.61 between D and D*GY-s to 0.95 between GY-s and D*GY-s.

 

Table 1. Estimates of heritability and genetic correlations for weight and backfat between purebred (PB) and crossbred (CB) populations (Merks and Hanenberg, 1998). * SE not estimable.

 

 

For the situation of the Stamboek breeding program, the use of PLS and reciprocal recurrent selection (RRS) may be applied in two ways:

 

1. Half sib RRS scheme (HS-RRS); selected pure line animals produce the next purebred generation and the crossbred progeny simultaneously.
2. Two stage RRS scheme (TS-RRS); two stage selection with independent culling levels for purebred and crossbred progeny performance. The purebred selection precedes each cycle of crossbred selection.

 

The results of the effectiveness for crossbred selection response of using only crossbred (RRS) versus only purebred (PLS) information are indicated in Table 2. For each sire 15 (half sibs) crossbred progeny are assumed.

 

Table 2. The effectiveness of using only purebred versus only crossbred information (PLS/RRS) for relative crossbred selection response in the system of half sib RRS (HS-RRS) and two stage RRS (TS-RRS) (Merks and Hanenberg, 1998)

 

 

The application of these results in a commercial pig breeding program is dependent on different factors. Two approaches basically may be used to achieve crossbred improvement. First, pure line selection (PLS) based on the breeding animal and its relatives’ information within populations is utilized. Under this approach, accomplished genetic gain in crossbreds arises from the regression on purebred response. Alternatively, reciprocal recurrent selection (RRS) is based on the breeding animals’ crossbred information. The last method has shown its advantage to improve traits with low heritability and large non-additive variation (Wei and Van der Steen, 1991).

The results of this study show the effectiveness of using crossbred data for genetic improvement of both purebreds and crossbreds. The effectiveness of CCPS or RRS versus PLS does not only depend on the genetic correlation between purebred and crossbred performance and crossbred heritability, but also on the chosen selection system. For traits with high heritabilities and high genetic correlations, the use of crossbred information is advantageous, only if gathering this information does not prolong generation interval.

2. Sire x herd interaction within the nucleus level (Merks, 1988)

In an earlier paper (Merks, 1987) the existence of sire x herd interaction in the nucleus level was reported for purebred on-farm test results. Merks (1988) investigated the significance of this sire x herd interaction, and the genetic correlations between the same traits measured across various herds were estimated.

 

Data used

The data used were from Dutch Landrace (NL) gilts and Dutch Yorkshire (GY) gilts and boars tested in the Stamboek field-testing program between May 1980 and December 1983. In the on-farm test, pigs are weighed (weight) and their backfat thickness is measured ultrasonically (backfat) at about 180 days of age. The study included data from 31,268 Dutch Landrace gilts and 16,828 Dutch Yorkshire boars.

 

Results

The sire x herd interaction was significant (P<0.001) for all traits in both data sets. The sire x herd interaction component may be inflated by differences in genetic scale between environments. Heterogeneity of genetic variances was investigated by estimating variance components within herds. Thirty to ninety percent of the sire x herd interaction was due to unequal genetic variances in the different herds. This heterogeneity of genetic variance was only related to the within herd standard deviation to a small extent, and even less to the herd level for weight or backfat.

The degree of relevance of the sire x herd interaction for the breeding program is measured by r_g. These correlations, presented in Table 3, may be underestimated due to genetic heterogeneity. To get an indication of the size of the possible underestimation, a correction with vâr (σ_Si) was performed. Also these adjusted intra-class correlations are tabulated in Table 3. The adjusted intra-class correlations (r’_g) are 25 – 50 % higher than the unadjusted correlations.

 

Table 3. Genetic correlations (r_g) between sires’ progeny in different herds and thegenetic correlations after correcting for genetic heterogeneity (r’_g) from Merks, 1988

 

 

Since the differences between herds are numerous and sometimes not definable (e.g. pathogen levels and management), it is very unlikely that only factors like feeding level or housing system are responsible for the sire x herd interactions. In that case selection of sires on the basis of sib or even progeny results in different herds may be desirable. Brascamp et al. (1985) indicated that progeny testing becomes more attractive than the use of sib results when the correlation between central test and on-farm test (indicated as r_g) is below 0.5. As the estimated correlations among herds approach the upper limit for r_g (Brascamp et al., 1985), at least for weight, sib and perhaps even progeny testing may be more efficient.

 

3. Purebred performance in SPF versus conventional health systems (Bergsma et al., 2001)

 

Breeding company TOPIGS has the availability of nucleus farms under conventional and under SPF health status. The question is whether or not selection will yield similar animals under both health conditions.

Data used

Genetic linkage between conventional and SPF herds was considered sufficient for an accurate estimate of the genetic correlation for reproduction parameters. These herds originated from the original conventional (Roosendaal) and the SPF nucleus (Brouennes) herd. Finishing traits, however, were hardly measured at the SPF foundation herd, but stem from one step further (Comigo and Kipling).

Bivariate genetic analyses were performed with a specific trait treated as two separate traits, one under conventional and one under SPF conditions. ASREML was used to calculate genetic parameters.

 

Results

Genetic parameters are given in Table 4. Two interesting results for reproduction traits: the genetic correlation between the two health environments was high not significantly different from 1 for litter size and #stillborn. It was high for gestation length, but somewhat lower than 1.0. Heritabilty for litter size was somewhat higher under SPF than under conventional, due to a higher genetic variance.

 

Table 4: Genetic parameters for reproduction and production traits under different health conditions (Bergsma et al., 2001)

 

	h2conv	h2 spf	p2 conv	p2 spf	rg	σ2 conv	σ2 spf
Litter size	0.099±0.011	0.135±0.020	0.0875±0.009	0.091±0.018	0.92±0.19	0.92	1.14
#stillborn	0.054±0.007	0.056±0.014	0.0493±0.008	0.059±0.015	1.00±0.28	0.11	0.10
Gest.length	0.317±0.016	0.309±0.028	0.086±0.011	0.080±0.021	0.76±0.13	0.72	1.00
Life gain	0.240±0.035	0.197±0.027	0.198±0.017	0.177±0.014	0.98±0.25	1032	779
Back fat	0.423±0.035	0.386±0.033	0.065±0.013	0.095±0.012	0.69±0.18	0.870	0.405

 

For life gain and for backfat somewhat higher heritabilities were estimated for the traits under conventional health. Reproduction information was available in a genetically well-structured data set. Genetic correlations for three different reproduction traits were hardly different from one. Heritability for litter size was somewhat higher under SPF than under conventional health conditions. For genetic improvement in the mentioned traits there is a slight advantage for selection under SPF, even for genetic progress under conventional situations.

Figure 2: Genetic correlations under different health conditions for litter size

4. Purebred performance in tropical versus temperate environment (Mote, 2000)

 

Bunge (BMI) has its nucleus farms both in temperate (Australia) and tropical (Indonesia) environment. By exchange of semen between these herds, Mote (2000) was enabled to find out how these conditions affected genetic progress.

 

Data used

Genetic parameters were estimated for growth rate from birth to slaughter age (GR), backfat measured at the P2 site (Ps) and number of live born piglets (NBA) in a temperate and a tropical environment. Semen was collected from 22 service sires (14 Large White and 8 Landrace) at Bunge in Australia (temperate environment) and part of this semen was sent to Culindo in Indonesia (tropical environment). This semen was used at both places, thus creating a genetic link between both operations.

Data used to calculate genetic parameters for GR and BF were collected for 19,304 progeny at BMI and 5,102 progeny at Culindo between July 1997 and August 1999. Farrowing records were collected from January 1997 to August 1999 and totalled 6,097 records at BMI and 4,083 records at Culindo.

 

Results

The heritability for GR was 0.25 for both environments. The heritability for P2 at BMI was 0.37 (±0.02), which was not significantly higher than the heritability of 0.30 (±0.04) found at Culindo. For NBA, the heritabilities were low in both environments (BMI: 0.08 ±0.02; Culindo: 0.03 ±0.02). The genetic correlation between GR and BF was –0.05 (±0.06) at BMI and was not significantly different from zero. At Culindo, this correlation was 0.16 (±0.11). Genetic correlations between NBA and performance traits ranged between 0.17 and 0.20 (±0.08-0.22).

The genetic correlation (r_g) estimated for GR between BMI and Culindo was 0.96 (±0.10). This indicated that a G x E interaction did not exist for GR (Table 5). For P2 the r_g was 0.78 (±0.16). The high standard error indicated that it was not significantly different from one. A G x E interaction was found for NBA with the r_g estimated as 0.31. Although this indicated that NBA was a different trait in the two operations, it is not very reliable given that it was not possible to estimate a standard error as a result of limited data for these traits.

 

Table 5. Correlations between BMI and Culindo for growth rate and backfat (Mote, 2000).

Abbreviations: GR – growth rate, P2 – backfat, LW – Large White, LR – Landrace, r_g (s.e.) – genetic correlations with standard error in brackets. ⁺ Standard error could not be obtained.

 

 

In conclusion, GR was the same trait in both operations. There was some uncertainty regarding a G x E interaction for P2 between the two piggeries. It is recommended to continue recording this trait under tropical conditions to monitor a G x E interaction. Data were insufficient to analyse a reliable estimate for a G x E interaction for NBA. Further data collection for NBA is required.

 

5. Purebred performance in nucleus herds versus commercial crossbred information (Van Steenbergen and Merks, 1998)

 

An important source of crossbred performance is of course that of commercial pigs.  Van Steenbergen and Merks (1998) presented the results of the estimation of r_pc and h² in the TOPIGS breeding program for a sire line and its crossbred offspring (fattening pigs). The estimated parameters were also used to investigate the relevance of using additional crossbred data for genetic progress in crossbred performance.

 

Data used    

Pure line performance data of the Great Yorkshire (GY-s) was used. The following traits were considered: ultrasonically measured backfat thickness (mm) and daily live weight gain (g/d). Data from 32,069 performance tested nucleus animals born between January 1993 and December 1996 were used in the analysis. Data were obtained from three nucleus farms. Since 1995 crossbred sows of two closed commercial herds were inseminated with semen of GY-s nucleus boars. Those boars were used simultaneously in nucleus and the commercial farms. All fattening pigs (gilts and castrates) were identified by means of an electronic ear button at an age of approximately 28 days. Data of 14,990 crossbred pigs were analyzed for the traits empty carcass weight over age (CDG) and backfat thickness (CBF) measured with the Henessy Grading Probe (HGP).

Crossbred animals descended from 127 GY-s boars, of which 85 also had performance tested offspring in the nucleus. 161 GY-s boars had offspring in the nucleus only. In the nucleus on average 60 offspring per sire were performance tested, while commercial data of over 110 animals per sire were available for analysis.

 

Results

The heritability of daily gain in crossbred pigs was shown to be higher than in pure line data, 0.37 and 0.28 respectively (Table 6). Daily gain for purebred pigs was estimated as live weight over age at a fixed age, while daily gain of crossbred pigs was based on differences in age at a more or less fixed carcass weight.

 

Table 6. Variance components and heritability estimates for daily gain and backfat thickness.

 

 

The genetic correlation between purebred and crossbred performance (r_pc) of daily gain and backfat thickness are 0.50 and 0.56 respectively (Table 7). The r_pc of daily gain is lower than accordance with results of Merks and Hanenberg (1998), who estimated r_pc 0.9-1.0 between test performance on purebred and crossbred boars in nucleus herds. The r_pc of backfat thickness is in better agreement with Merks and Hanenberg (1998), 0.56 versus 0.6-0.7 respectively.

 

Table 7. Genetic correlations between daily gain and backfat thickness

 

 

Estimated (co)variances were used to calculate genetic progress for daily gain and backfat thickness of crossbred pigs. In Table 8 the ratio genetic progress CCPS/PLS with various numbers of crossbred offspring per sire is given per trait.

 

Table 8. Ratio of CCPS to PLS response in crossbred performance with various numbers of crossbred offspring per sire

 

 

Assuming a breeding structure where nucleus sires are used simultaneously in the nucleus and on commercial farms, an extra genetic response for crossbred daily gain and backfat thickness can be achieved of 53 % and 32 % respectively. This extra selection response is even higher than selection response based on BLUP Animal model compared to selection on own performance. When using CCPS, one should be aware that genetic progress of daily gain and backfat thickness in the purebred sire line will decrease by 32 % and 20 %.

 

6. Purebred versus crossbred performance for litter size (Täubert et al., 1998)

 

In most crossbreeding programs the selection for purebred performance in the nucleus is the method of choice because the selection for crossbred performance is very expensive and takes a lot of time. The development of the animal model with the use of a relationship matrix made it possible to build a model for the estimation of the genetic correlation between purebred and crossbred performance. The performance of purebred and crossbred animals is treated as different traits and the analysis is made under a multiple trait animal model.

 

Data used

For the analysis two data sets were available. The first data set was an Australian data set with 8,341 Large White sows, 3,236 Landrace sows and 12,127 Large White x Landrace crossbred sows. Both reciprocal crosses were available. The second data set was taken from the German breeding company Bundeshybridzuchtprogramm (BHZP). Records were available from the multiplier herds for 11,577 purebred sows and from sow management programs for 12,127 crossbred sows. The sows from the multiplier herds could be identified as dams of the crossbred sows, so the relationship between purebred and crossbred sows could be established. Because of the missing reciprocal cross only one purebred dam line and the F1 daughters could be analyzed.

 

Results

The results of the estimated heritabilities and genetic correlations between purebred and crossbred performance are shown in Table 9 (Australian data set) and in Table 10 (German data set).

 

Table 9. Heritabilities and genetic correlations between purebred and crossbred litters in the Australian data set

 

 

Table 10. Heritabilities and genetic correlations between purebred and crossbred litters in the German data set

 

 

                Heritabilities in both data sets are in the expected range of 0.05 – 0.10. The standard errors for the heritabilities of the purebred sows are in the range between 0.0044 and 0.0066, for the heritabilities of the crossbred sows in the range between 0.01 and 0.12 and for the genetic correlations in the range between 0.003 and 0.120. The heritabilities in the Australian data set are very similar to those in the German data set. In the Australian data set the second litter of the purebred sows has the highest heritability of the three litters, whereas in the German data set the heritability is continuously rising with the number of litters. Both have been noticed in former analyses (Roehe and Kennedy, 1995). In crossbred parameters the second litter shows the smallest heritability in both data sets, where in the German data set it nearly amounts to the half of the other two litters. The genetic correlations between purebred and crossbred performance are similar in both data sets and very high. They are, except the value for the second litter in the Australian data set, not significantly different from 1.

 

Implications

Finally, the consequences of the moderate genetic correlations for the design and efficiency of pig breeding programs need to be considered. In general, the accuracy of selection across levels of the breeding program is directly proportional to (r_G / [ r_gP * r_gH ] ^½ )  (Merks, 1988) where r_gP and r_gH are the genetic correlations within respectively the level where the performance information is collected and the level where the breeding goal is defined. However, for a fixed r_G, the greatest genetic progress may be reached if r_gP and r_gH are the genetic correlations within the level where the index information is collected and the level where the breeding goals are defined, respectively. However, for a fixed r_G, the greatest genetic progress may be reached if r_gP = r_G. A limited number of test places are best used by distributing the representatives of the genotype over as many herds as possible.

Results of the different G x E studies summarised in this paper indicate poor genetic correlations for specific within farm differences (0.7 – 1.0), general across farm differences (0.5 – 1.0), pure versus crossbreeding nucleus (0.6 – 1.0), and consequently for nucleus versus commercial level (0.5 – 1.0). However, it should be mentioned, that traits measured are not always identical, that the connectedness of data is far from optimal.

            For improved genetic progress at the commercial level, in general, the options available are 1) using purebreds at the commercial level and 2) making the nucleus level more comparable with the commercial level (traits, management etc.) However, both options are not practical. The options applied by TOPIGS are 1) direct supply of genetics from the nucleus, 2) using the satellite nucleus system across environments, and 3) applying Combined Crossbred and Purebred Selection (CCPS).

 

Figure 3. Satellite Nucleus System TOPIGS

In the nucleus, the combination of purebred selection and selection on crossbred information only, it is important that the use of crossbred information should NOT prolong generation interval, i.e. AI boars should produce purebreds and crossbreds simultaneously. The effect largely depends on r_gpc and h² of both purebred and crossbred performance. The effects calculated range from 0 to >100 % extra genetic progress. In the nucleus, genetic progress is lower when CCPS is applied.

 

Conclusions

In most breeding programs genetic progress is lower at the commercial level than expected from genetic progress at the nucleus level:

-          Traits measured are not identical (e.g. ultrasonic backfat versus carcass backfat).

-          Variation within and between farm environments causes genotype x environment interaction.

-          Progress in pure line breeding is not equal to genetic progress in crossbreeding systems (measured by progeny of same sires).

Genetic progress at the commercial level can be improved very much by:

-          The definition of the breeding goal at the commercial level.

-          Application of the satellite nucleus system across various ‘environments’.

-          Application of the Combined Crossbred and Purebred Selection (CCPS) system.

 

References

Bergsma, R., Knol, E.F., Merks, J.W.M. and Van Groenland, G.J. 2001. Performance levels, genetic parameters and genotype-health interactions for production traits in pigs. American Society for Animal Science. Anim. Sci. 79 (suppl. 1): 186 (769).

 

Bijma, P., Van Arendonk, J.A.M. 1998. Maximizing genetic gain for the sire line of a crossbreeding scheme utilizing both purebred and crossbred information. Anim. Sci. 66: 529-542.

 

Brascamp, E.W., Merks, J.W.M. and Wilmink, J.B.M. 1985. Genotype-environment interaction in pig breeding programs: methods of estimation and relevance of the estimates. Livest. Prod. Sci., 13:135-146.

 

Falconer, D.S. 1952. The problem of environment and selection. Amer. Nat., 86:293-298.

 

Merks, J.W.M. 1988. Genotype x environment interactions in pig breeding programs. Thesis. Wageningen Agricultural University. The Netherlands. 137pp.

 

Merks, J.W.M. and Hanenberg, E.H.A.T. 1998. Optimal selection strategy for crossbred performance in commercial pig breeding programs. Proceedings 6^th World Congress on Genetics Applied to Livestock Production. Armidale, Australia. 23: 575-579.

 

Mote, T.G. 2000. Genotype by environment interactions between pig populations in Australia and Indonesia. Thesis. University  of New England, Australia. 96pp.

 

Roehe, R. and Kennedy, B.W. 1995. Estimation of genetic parameters for litter size in Canadian Yorkshire and Landrace Swine with each parity of farrowing treated as a different trait. J. Anim. Sci. 73: 2959-2970.

 

Täubert, H., Brandt, H. and Glodek, P. 1998. Estimation of genetic parameters for farrowing traits in purebred and crossbred sows and estimation of their genetic relationships. Proceedings 6^th World Congress on Genetics Applied to Livestock Production. Armidale, Australia. 23: 597-582.

 

Van der Werf, J.H.J., Van der Wei, M. and Brascamp, E.W. 1994. Combined crossbred and purebred selection to maximize genetic response in crossbreds. Proceedings 5^th World Congress on Genetics Applied to Livestock Production. 18: 266-269.

 

Van Oijen, M.A.A.J. and Merks, J.W.M. 1995. Estimates of genotype x environment interaction and their impact on pig breeding programs. 46^th Annual Meeting European Association for Animal Production. Prague, Czech Republic. 6pp.

 

Van Steenbergen, E. and  Merks, J.W.M. 1998. Use of slaughterhouse information of crossbred fattening pigs in pure line selection. 49^th Annual Meeting European Association for Animal Production. Warsaw, Poland. 4pp.

 

Wei, M, and Van der Steen, H.A.M. 1991. Comparison of reciprocal recurrent selection with pure-line selection systems in animal breeding (a review). Anim. Breed. Abstr. 59: 281-298.