If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Department of Agronomy, Food, Natural Resources, Animals and Environment (DAFNAE), University of Padova, Viale dell' Università 16, 35020, Legnaro PD, Italy
Unidad de Genética Cuantitativa y Mejora Animal, Instituto Agroalimentario de Aragón (IA2), Universidad de Zaragoza, Calle de Miguel Servet, 177, 50013, Zaragoza, Zaragoza, Spain
Department of Agronomy, Food, Natural Resources, Animals and Environment (DAFNAE), University of Padova, Viale dell' Università 16, 35020, Legnaro PD, Italy
Department of Agronomy, Food, Natural Resources, Animals and Environment (DAFNAE), University of Padova, Viale dell' Università 16, 35020, Legnaro PD, Italy
Our study investigated the inbreeding load for fertility traits in the Italian Brown Swiss dairy cattle breed. Fertility traits included continuous traits (i.e., interval from calving to first service, days open, and calving interval) and categorical traits (i.e., calving rate at first insemination and nonreturn date at d 56). We included only records of the first 3 parities of cows that calved between 2010 and 2018. We traced up the pedigree of the cows with records as far as possible, ending up with a total of 73,246 animals. The final data set consisted of 59,864 records from 34,921 cows. We analyzed all models using a Bayesian approach that included a covariate with total inbreeding in addition to systematic, permanent environment, additive genetic, and inbreeding load effects. We then evaluated the trends in heritabilities and ratios of the inbreeding load using a continuum of partial inbreeding coefficients from 0.001 to 0.100 as reference. Posterior estimates of heritabilities tended to decrease across the continuum, whereas ratios of the inbreeding load tended to increase, more noticeably in categorical traits (calving rate at first insemination and nonreturn date at d 56). From the results obtained, we confirmed the presence of heterogeneity in inbreeding depression. We then predicted the inbreeding load effects, which had a low reliability of prediction, explained by having only 513 ancestors generating inbreeding. However, reliability of prediction was high enough for some of the individuals, obtaining a favorable prediction of inbreeding load for a relevant percentage, which improved the phenotypic performance of their inbred descendants. These results make it feasible to implement breeding and management strategies that select ancestors with a favorable inbreeding load prediction. In addition, it opens the possibility to define a global index for the expected consequences of the inbreeding generated by each individual.
). It increases the probability of recessive alleles in homozygosis and decreases the percentage of heterozygous loci, thereby losing the advantage of overdominance (
Biffani, S., A. B. Samoré, and F. Canavesi. 2002. Inbreeding depression for production, reproduction and functional traits in Italian Holstein cattle. 7th World Congress on Genetics Applied to Livestock Production. Communication No. 09-44.
Inbreeding depression is a genetic phenomenon. It depends on the genotype of the ancestors whose alleles produce identity by descent in their progeny (
Miglior, F., E. B. Burnside, and W. D. Hohenboken. 1994. Heterogeneity among families of Holstein cattle in inbreeding depression for production traits. 5th World Congress on Genetics Applied to Livestock Production. Vol. 28, pages 479–482, Guelph, Ontario, Canada.
suggested a Mendelian decomposition of inbreeding that split inbreeding among founders and the Mendelian sampling of the nonfounders. This Mendelian decomposition was the basis on which
proposed a new parameterization for predicting inbreeding loads for the individuals generating inbreeding and those not generating it. The availability of prediction of inbreeding loads allows us to distinguish between favorable and unfavorable effects on a specific trait. This information could help develop optimal breeding strategies whereby individuals able to provide favorable inbreeding effects on a trait (e.g., decreasing the number of days open) are mated (
Man, W. Y. N., J. W. James, and F. W. Nicholas. 2002. Effect of inbreeding contribution from particular ancestors: A preliminary analysis of first lactation milk yields from Holstein Friesians in Australia. 7th World Congress on Genetics Applied to Livestock Production. Communication No. 23-01.
). It could also serve as a tool for artificially purging individuals carrying deleterious alleles, ensuring that individuals with an unfavorable inbreeding load effect are not used as reproducers (
, we conducted a study with the objectives of (1) estimating the genetic parameters and inbreeding load variances using a continuum of partial inbreeding coefficient as a reference and (2) predicting the inbreeding load effects on fertility traits in the Brown Swiss dairy cattle breed.
MATERIALS AND METHODS
Data Set
The data for the study consisted of records of the fertility traits of Brown Swiss cattle collected by the Breeders Federation of Alto Adige/Südtirol (Associazione Provinciale delle Organizzazioni Zootecniche Altoatesine/Vereinigung der Südtiroler Tierzuchtverbände, Bolzano/Bozen, Italy) in northeastern Italy. Fertility traits included continuous traits [interval from calving to first service (ICF), days open (DO), and calving interval (CInt)] and categorical traits [calving rate at first insemination (CR) and nonreturn rate at d 56 (NR56)]. The ICF and DO records were considered censored if the cow's pregnancy was not confirmed. For DO and ICF, we discarded records covering periods of less than 20 d. For CR, 1 referred to a cow being pregnant, and for NR56, 1 referred to a cow with no second insemination registered and an observation interval of >56 d. A detailed description of the data set is given in
. A total of 49,184 cows were available, with 234,877 phenotypic records of all fertility traits before data editing. We included only records regarding the first 3 parities of cows that calved between 2010 and 2018. The original pedigree had 2,793,159 animals, and we extracted the subpedigree, including all animals with records and all their known ancestors, ending up with a total of 73,246 animals. After editing, 21,921 to 34,921 cows with phenotypic information remained, with between 29,860 and 59,864 phenotypic records. The average (±standard deviation) phenotypes of the continuous fertility traits were 91.8 d (±41.0) for ICF, 127.8 d (±70.2) for DO, and 419.5 d (±75.4) for CInt, with an incidence of censored records of 20.3% for ICF and 24.2% for DO. The incidence of categorical traits was 49.9% for CR and 33.1% for NR56 (see Table 1).
Table 1Mean (SD in parentheses) of continuous traits and incidence (%) of categorical traits in addition to minimum, maximum, and percentage of censored records in Brown Swiss cows after editing
where y is the vector of phenotypic records of the continuous traits (ICF, DO, and CInt) or the liabilities of the categorical traits (CR and NR56); b is the vector of systematic effects: parity and year-season of calving (a combination of the year and the season of calving in 12 categories of 5 d each; season 1 = calvings from April to September, season 2 = October to March); h, p, a, i, and e are the vectors of random effects of herd, permanent environment, animal additive genetic, inbreeding load, and residuals, respectively; f is the vector of total inbreeding of the recorded individual; c is the covariate with total inbreeding; a and i are genetic effects: a (the additive genetic effect) is expressed in all phenotypes, whereas i (the inbreeding load) is expressed in the phenotypes of the inbred descendants (
); and X, W1, W2, Z, and K are the incidence matrices corresponding to the vectors of systematic, herd, permanent environment, additive genetic, and inbreeding load effects, respectively. Following transformation of the partial inbreeding matrix described by
, K = T(I − P), where T is a lower triangular matrix with each of the nonzero elements corresponding to the partial inbreeding coefficient (Fp) linking the phenotype of an inbred individual with the ancestor causing inbreeding. The Fp were obtained by Mendelian decomposition of inbreeding, following the procedure of
. For computational reasons, we then multiplied by 10 to obtain the inbreeding load variance for an Fp from one ancestor of 0.10. P is a projection matrix with 0 in the diagonal and 0.5 in the elements that link individuals with its sire and dam, and I is the identity matrix. A detailed description of the procedure for calculating the Fp and an R script is presented in Supplemental Files S1 and S2 (https://figshare.com/s/f1d2e7f1f7a71355170c).
Under a hierarchical Bayesian scheme, it was assumed that prior distributions for the herd, permanent environment, and residual effects had multivariate Gaussian distributions:
where
,
, and
are the herd, permanent environment, and residual variances, respectively. In addition, the prior distribution of the additive genetic and inbreeding load effects was
where A is the numerator relationships matrix, and
where
and
are the additive genetic variance, the inbreeding load variance, and the covariance between the additive genetic and the inbreeding load effects, respectively. Finally, the prior distribution of the systematic effects (b) and the variance components
was uniform within appropriate bounds.
Gibbs Sampler
We analyzed all models using a standard Bayesian approach and estimated the marginal posterior distributions of unknown parameters by Gibbs sampler using software developed by
. The total number of iterations was 2,500,000, with a burn-in of 500,000 and without thinning. Analysis of the inbreeding coefficients (F) for the individuals in the pedigree was carried out in the INBUPGF90 program of the family of BLUPF90 software programs (
), which was also used for the post-Gibbs analyses and for estimating the correlations between additive genetic and inbreeding load [r(a,i)], heritabilities (h2), and the ratios of the inbreeding load (i2), herd (he2), and permanent environment (pe2) variances, and the slope of c (β). The r(a,i) was defined as
where σai is the covariance between the additive genetic and the inbreeding load, and
and
are the additive genetic and inbreeding load variances, respectively. The i2 was defined as the proportion of phenotypic variation caused by variability in the inbreeding load variance. The calculation of i2 requires the assumption of a reference point defined by the inbreeding caused by one specific ancestor. We used a continuum from F = 0.001 to F = 0.100 in sequences of 0.001 to reflect the trend associated with i2 in different scenarios to an extreme (F = 0.100) with highly inbred individuals. To estimate i2 across the continuum, we rescaled the estimates depending on the reference point of F by multiplying it by the square of the ratio between the 2 reference points (the original and the new one). Thus, the posterior estimate of the inbreeding load for ICF with a reference point of F = 0.025 and based on F = 0.100 was 30.842 = 493.471 × (0.025/0.100)2. We then calculated the corresponding h2 that reflects the change in the magnitude due to i2. The estimates of β were defined as the deviations from F. Furthermore, the reliability of the predictions of the jth inbreeding load was approximated as
where PSD(ij) is the posterior standard deviation of the jth inbreeding load, and
is the posterior mean estimate of the inbreeding load variance.
RESULTS AND DISCUSSION
Mendelian Decomposition of Inbreeding
The fraction of inbred animals from the original pedigree (2,793,159 individuals) was 59.6%, with an average F = 0.020 (±0.031). After editing (73,246 individuals), the incidence of inbred animals was 94.6%, and F ranged from 0.031 (25th percentile) to 0.058 (75th percentile), 0.071 (90th percentile), and 0.080 (95th percentile), with an average of 0.045 (±0.023). The average F was twice as high in the edited pedigree as in the original pedigree. This difference can be explained by the loss of information in the edited pedigree, as the records were obtained from a group of farms. The majority of the genealogy is not connected with the phenotypic records.
Mendelian decomposition of inbreeding from the edited pedigree generated 268,525 Fp and an average of 3.66 ancestors whose alleles can generate identity by descent. The average Fp was 0.0023, with a standard deviation of 0.003. As shown in Table 2, the distribution was asymmetric, with 7,518 Fp greater than 0.01 (2.80% of the total) and only 162 greater than 0.1 (0.06% of the total). The maximum Fp was 0.125. These results indicate that most of the ancestors generating inbreeding are from several generations ago, whereas recent ancestors generated only a small percentage of inbreeding. Moreover, as shown in Table 2, these Fp were generated by only 513 ancestors. Of these, 188 individuals generated inbreeding to more than 10 individuals, 93 to more than 100, 29 to more than 1,000, and 9 to more than 10,000. As the 513 ancestors are genetically linked to all individuals in the pedigree, the i of the remaining individuals can be predicted based on the assumption of variation in the additive genetic effect of the i and the genetic relationships between the individuals captured by the A matrix.
Table 2Number of animals, sires, dams, and generations from the edited pedigree,
The posterior mean (and posterior standard deviation) estimates of variance components are presented in Table 3. The slope of the covariate with the total F was positive for continuous traits, ranging from 0.260 (−7.30; 7.67) in CInt to 8.43 (0.49; 16.45) in ICF, and negative for categorical traits (~0.078). The most important source of phenotypic variation in all fertility traits was the residual variance, ranging from 1,859.0 (1,793.2; 1,927.1) days squared in ICF to 6,485.5 (6,346.8; 6,627.5) days squared in DO for continuous traits, and was set to 1 for the categorical traits (CR and NR56). The inbreeding load variances were larger than the additive genetic variances, more noticeably in categorical traits [0.021 (0.013; 0.032) vs. 0.123 (0.025; 0.374) in CR and 0.019 (0.010; 0.030) vs. 0.146 (0.037; 0.382) in NR56], except for trait DO, where the estimate of genetic variance was larger [703.4 (546.4; 874.5) vs. 629.6 (159.0; 1,926.5)]. The posterior estimates of the genetic correlation between the additive and the inbreeding load effects were negative, ranging from −0.141 (−0.751; 0.639) in CInt to −0.281 (−0.793; 0.445) in ICF, and close to zero in DO −0.029 (−0.775; 0.734), whereas the HPD95 intervals included zero in all traits. In their study,
obtained a negative correlation for weaning weight in the Pirenaica breed and a correlation close to zero in the Rubia Gallega breed and suggested that the different results could be due to the differences in the depths of the pedigrees available for each breed. In our final data set, the pedigree depth was smaller than that of the original pedigree (13 vs. 39 generations). Only a small number of ancestors generated inbreeding (513 individuals), which could explain the need for more information to estimate the correlations more accurately.
β(F = 0.10) = slope of the covariate with total inbreeding; σa2 = additive genetic variance; σi(F=0.10)2 = inbreeding load variance at a partial inbreeding coefficient of 0.10; σia(F=0.10) = covariance between the additive genetic and inbreeding load effects at a partial inbreeding coefficient of 0.10; r(a,i) = correlation between additive genetic and inbreeding load, r(a,i)=σaiσa2×σi2;σh2 = herd variance; σpe2 = permanent environment variance; σe2 = residual variance.
ICF
DO
CInt
CR
NR56
β(F = 0.10)
4.54 (−1.84; 10.79)
8.43 (0.49; 16.45)
0.26 (−7.30; 7.67)
−0.068 (−0.169; 0.039)
−0.087 (−0.191; 0.022)
244.9 (177.4; 322.9)
703.4 (546.4; 874.5)
309.6 (229.3;402.7)
0.021 (0.013; 0.032)
0.019 (0.010; 0.030)
493.5 (149.4; 1213.1)
629.6 (159.0; 1926.5)
770.1 (148.3; 2303.3)
0.123 (0.025; 0.374)
0.146 (0.037; 0.382)
−103.9 (−369.3; 130.1)
−27.5 (−657.7; 529.4)
−72.8 (−448.3; 315.6)
−0.007 (−0.039; 0.027)
−0.013 (−0.005; 0.027)
r(a,i)
−0.281 (−0.793; 0.445)
−0.029 (−0.775; 0.734)
−0.141 (−0.751; 0.639)
−0.157 (−0.706; 0.533)
−0.238 (−0.735; 0.565)
646.6 (584.1; 713.8)
1,046.3 (942.6; 1,157.7)
554.3 (495.9; 617.0)
0.085 (0.074; 0.095)
0.098 (0.087; 0.111)
447.6 (360.9; 532.9)
1,278.9 (1,099.6; 1,457.3)
539.0 (438.1; 638.6)
0.054 (0.033; 0.076)
0.042 (0.019; 0.066)
1,859.0 (1,793.2; 1,927.1)
6,485.5 (6,346.8; 6,627.5)
4,343.6 (4,255.5; 4,432.7)
1
1
1 HPD95 = lower and upper bounds of the 95% highest posterior density region.
2 ICF = interval from calving to first service; DO = days open; CInt = calving interval; CR = calving rate; NR56 = nonreturn rate at d 56.
3 β(F = 0.10) = slope of the covariate with total inbreeding; = additive genetic variance; = inbreeding load variance at a partial inbreeding coefficient of 0.10; = covariance between the additive genetic and inbreeding load effects at a partial inbreeding coefficient of 0.10; r(a,i) = correlation between additive genetic and inbreeding load, = herd variance; = permanent environment variance; = residual variance.
The relative magnitudes of the variance components are expressed by the heritabilities (h2) and the ratios of the inbreeding load (i2). These estimates, which depend on the amount of inbreeding generated by specific ancestors, are presented in Figure 1 under a continuum of different scenarios with different assigned Fp. The scenarios represent theoretical populations where all individuals are assigned Fp = 0.001 to 0.100. Posterior estimates of i2 tended to increase when the Fp assigned increased up to 0.10, increasing up to around ~0.12 in all traits except for DO, which increased around half as much (~0.06). Conversely, posterior estimates of h2 tended to decrease when the Fp assigned increased up to 0.10, although not as much as i2 increased. In continuous traits, the decrease of h2 was more pronounced compared with categorical traits. In continuous traits, h2 decreased in 1, 0.6, and 0.5% in traits ICF, DO, and CInt, respectively, whereas in categorical traits it decreased in only 0.2%. The changes in the estimates of h2 are due to the additional variance component, the i2, contributing to the phenotypic variation, more pronounced as the reference point increased and in categorical traits.
Figure 1Trend of the posterior estimates of the inbreeding load variance ratio (i2; continuous line) and heritabilities (h2; dashed line) across a continuum of partial inbreeding coefficients.
The results obtained for i2 should be interpreted as the variation for one unit explained by the heterogeneity of the inbreeding depression effects in a theoretical population where each of the individuals has an Fp generated by a single, specific ancestor. Alternatively, it can be understood as the additional randomness in an inbred individual's phenotypic performance with a given probability of identity by descent from a single ancestor. However, large Fp are infrequent in commercial farms that use breeding strategies that try to avoid recent inbreeding, as in the pedigree we analyzed. The estimates of the i2 indicate that the heterogeneity of inbreeding depression had an almost negligible effect on our phenotypes. Nonetheless, we were able to confirm the presence of heterogeneity of inbreeding depression suggested in previous studies (
Miglior, F., E. B. Burnside, and W. D. Hohenboken. 1994. Heterogeneity among families of Holstein cattle in inbreeding depression for production traits. 5th World Congress on Genetics Applied to Livestock Production. Vol. 28, pages 479–482, Guelph, Ontario, Canada.
), although its effects should be noticeable only in the phenotypic variation in individuals with a large amount of inbreeding. However, the i2 we obtained were higher than those obtained by
). The inbreeding depression we obtained could be due to the presence of recessive semideleterious alleles that negatively affect fertility, as seen in other studies (
). Moreover, the distribution of these recessive alleles may be heterogeneous in individuals in the population and may determine the variability in individual inbreeding loads.
provides predictions of the inbreeding loads (i) for all of the individuals in the pedigree. However, in our study, the information available for predicting i of all individuals was associated with the genetic link to one of the 513 individuals generating Fp. The average reliability [r(ij)] was low, as shown in Table 4, ranging from 0.122 in CR to 0.179 in ICF. Given the low r(ij), most of the predictions of i were very close to zero. However, a relevant fraction of the individuals had higher r(ij) because they generate inbreeding in the phenotyped individuals or have a strong genetic link to them. Using as a threshold an r(ij) of 0.3, the number of individuals with a greater value ranged from 3,901 in CR to 7,233 in DO. The distribution of the predicted i was variable and centered to zero, as shown in Figure 2. The slope with the covariate of the total F (β) indicates the average effect of inbreeding, added for visualization purposes. That is, when i> or <β, the effect of inbreeding depression will be either positive or negative. The β obtained was positive for continuous traits, highest in DO (8.43), and close to zero in CInt (0.26), indicating a worsening in days in the phenotypic performance of the descendants due to inbreeding depression. In addition, it was negative for categorical traits (~0.029), indicating a worsening in percentage in the phenotypic performance of the descendants. However, a percentage of individuals had predictions of i with favorable effects with respect to β, improving the phenotypic performance of their inbred descendants by reducing the number of days in continuous traits and increasing the incidence in categorical traits. These results were most noticeable for CInt (53.3%) and ICF (17.1%) and least noticeable for DO (0.2%). These variabilities can also be seen in Table 5, which presents the predicted i for the 5 individuals with the highest r(ij), which correspond to old AI sires that generate inbreeding in a large number of individuals (370–15,272) and have a vast number of sons (53–1,930), grandsons (873–3,668), and great-grandsons (2,917–13,762). Availability of the predicted i opens new possibilities toward developing population breeding and management strategies instead of the general strategy of avoiding or limiting inbreeding (
). Although it is highly recommended to avoid inbreeding generated by a common ancestor with an unfavorable prediction of inbreeding load (worsening the phenotype), those ancestors with a favorable inbreeding load prediction could be allowed—or even favored—in breeding strategies. Moreover, inbreeding cannot be completely eradicated in some populations due to their limited effective size. Therefore, an alternative strategy may include predicting the inbreeding loads in the selection index to generate a selection response that mimics and accelerates purging effects (
). However, implementing these approaches involves predicting inbreeding loads for all traits of interest, and these traits may be genetically correlated. In fact, the raw correlations among the predicted inbreeding loads for all individuals ranged up to 0.817 between CR and NR56, suggesting the possibility of the development of a multivariate approach, which could be of interest in future studies.
Table 4Estimates of reliability (mean and SD) for each fertility trait, number of individuals with a reliability greater than 0.3, 0.4, 0.5, and 0.7, and maximum value of reliability in Brown Swiss cows
Figure 2Predicted inbreeding load (i) for continuous (ICF, DO, and CInt, in days) and categorical (CR and NR56, in %) fertility traits, the slope with the covariate of the total inbreeding coefficient (F; β, dashed line), and the proportion of the predicted i with a positive effect with respect to the slope (gray area). ICF = interval from calving to first service; DO = days open; CInt = calving interval; CR = calving rate; NR56 = nonreturn rate at d 56.
ICF = interval from calving to first service; DO = days open; CInt = calving interval; CR = calving rate; NR56 = nonreturn rate at d 56; F generated = number of individuals from which inbreeding is generated; Fp = partial inbreeding coefficient.
Individual ID
12783
12782
21519
12530
11170
Year of birth
2006
—
1995
—
—
ICF (d; reliability)
−22.32 (0.79)
8.07 (0.82)
22.12 (0.72)
1.18 (0.66)
7.41 (0.63)
DO (d; reliability)
−16.20 (0.72)
−1.46 (0.76)
10.81 (0.73)
4.88 (0.49)
0.17 (0.64)
CInt (d; reliability)
−13.12 (0.80)
2.46 (0.86)
12.77 (0.75)
−3.62 (0.68)
−12.86 (0.69)
CR (%)
0.027 (0.77)
−0.199 (0.76)
−0.013 (0.73)
−0.054 (0.67)
−0.204 (0.67)
NR56 (%)
−0.067 (0.78)
−0.286 (0.75)
0.106 (0.75)
−0.082 (0.70)
−0.087 (0.67)
Sons (no.)
251
403
1,930
385
53
Grandsons (no.)
3,548
3,583
3,688
3,290
873
Great-grandsons (no.)
13,762
8,636
2,917
8,772
7,167
F generated (no.)
15,272
2,552
370
3,873
11,170
Fp
0.00614
0.0077
0.0201
0.0061
0.0027
1 ICF = interval from calving to first service; DO = days open; CInt = calving interval; CR = calving rate; NR56 = nonreturn rate at d 56; F generated = number of individuals from which inbreeding is generated; Fp = partial inbreeding coefficient.
This study confirms the presence of heterogeneity in the inbreeding depression effects, reflected in the variance in the inbreeding load effects. It also confirms the presence of a new source of phenotypic variation—the inbreeding load variance—when inbreeding is present. The results confirm the strong influence of inbreeding depression on fertility traits and the ability to predict the inbreeding loads of individuals. The study opens the possibility to develop alternative breeding strategies that include individuals with a favorable inbreeding load, which can improve the descendants' phenotype (e.g., shortening DO or increasing CR). In light of this, the results could be used as a reference for creating a selection index in breeding strategies to obtain a response to selection by including traits of interest and predicting their inbreeding loads.
ACKNOWLEDGMENTS
We thank the Italian Brown Swiss Cattle Breeders Association (ANARB, Verona, Italy) for providing pedigree information. The research was part of the project ECOlatte (funded under the National Rural Development Program–Sottomisura 10.2: Animal Biodiversity; Italy). The authors have not stated any conflicts of interest.
REFERENCES
Adamec V.
Cassell B.G.
Smith E.P.
Pearson R.E.
Effects of inbreeding in the dam on dystocia and stillbirths in US Holsteins.
Biffani, S., A. B. Samoré, and F. Canavesi. 2002. Inbreeding depression for production, reproduction and functional traits in Italian Holstein cattle. 7th World Congress on Genetics Applied to Livestock Production. Communication No. 09-44.
Man, W. Y. N., J. W. James, and F. W. Nicholas. 2002. Effect of inbreeding contribution from particular ancestors: A preliminary analysis of first lactation milk yields from Holstein Friesians in Australia. 7th World Congress on Genetics Applied to Livestock Production. Communication No. 23-01.
Miglior, F., E. B. Burnside, and W. D. Hohenboken. 1994. Heterogeneity among families of Holstein cattle in inbreeding depression for production traits. 5th World Congress on Genetics Applied to Livestock Production. Vol. 28, pages 479–482, Guelph, Ontario, Canada.
00681-0&id=gr1.jpg){kind=link}
00681-0&id=gr2.jpg){kind=link}