All experimental data and pedigree records were obtained from Inner Mongolia Sano Sheep Breeding, including 2,022 early growth trait records of Dorper sheep collected from 2019 to 2021. The dataset was rigorously preprocessed in Excel: records with issues like incorrect ear tags, missing gender or production performance data, incomplete phenotypic measurements, and outliers outside “mean±3σ” were excluded, resulting in 1,999 lambing records involving 14 sires and 328 dams. The study focused on four early growth traits—BW, WW (adjusted to 90 days due to varying weaning ages), pre-weaning ADG, and pre-weaning KR. The KR is an indirect selection method used to measure feed conversion. The KR before weaning is calculated as follows [17]:

Where, W1 = WW, D = number of days between weaning date and date of birth.

The experimental population was housed at the company’s production base situated in Ulanqab (41°10′−43°22′N, 110°20′−113°00′E), a semi-arid continental monsoon climate zone with 313.8 mm annual precipitation (60%–70% in summer) and summer temperatures averaging below 20°C.

The sheep were reproduced by artificial insemination and LOPU-IVF-ET at 18 months of age. Newborn lambs were ear-tagged and weighed within 24 hours post-partum, with dam-suckling permitted during grazing until 30 days old. From 2 months postpartum, lambs received twice-daily suckling sessions (morning/evening) alongside daytime hay supplementation until weaning. Post-weaning lambs were transitioned to unified feeding regimens in grouped pens.

The fixed factors, including the birth year (2019–2021), birth month (ten months), herd (ten herds), sex (male or female), and age of recipient dam (2–4 years old), were analyzed using a General Linear Model (GLM) procedure using R software ver. 4.3.1. Post-hoc pairwise comparisons were conducted using Duncan’s multiple range test to determine homogeneous subsets among significantly different factor levels. A statistical model of GLM for the trait of per-weaning is described below:

Where y_ijklmr is the individual observation, μ is the overall mean, Y_i is fixed effect of birth year, H_j is fixed effect of herd, M_k is fixed of birth month, S_l is fixed of sex, and A_m is fixed of age of recipient dam, and e_ijklmr is the residual error.

Genetic parameters of pre-weaning growth traits were estimated by the AI-REML method using the ASReml ver. 4.1 [18]. Maternal permanent environmental effects (Model 2) were included to account for consistent non-genetic influences from recipient dams (e.g., milk quality, rearing behavior) across multiple offspring, a critical factor in embryo-transferred populations. Finally, we calculated the fitting parameters for six single-trait animal models and determined the optimal model for each trait using Akaike information criterion (AIC), Bayesian information criterion (BIC), and likelihood ratio test (LRT) tests. We estimated heritability using single-trait models, calculated genetic correlations between trait combinations using bivariate analysis, and then estimated heritability. The animal models are as follows:

Where y is a vector of observed traits; b, a˜N(0, Aσa2),m˜N(0, Aσm2), c˜N(0, Idσa2), are vectors of fixed, direct additive genetic, maternal additive genetic, maternal permanent environmental effects, respectively. e˜N(0, Inσe2) is a vector of residual effect; X, Z₁, Z₂, and Z₃ are correlation structure matrices for these effects relating b, a, m, and c to y. It was assumed as follows [19]:

A represent the additive genetic correlation matrix, and I_d and I_n are identity matrices of records equal to the number of dams and lambs, respectively. All components with the phenotypic variance ( σp2) being the sum of σa2,σm2,σc2, and σe2 were derived at convergence.

The direct heritability:

The maternal heritability:

The proportion of phenotypic variance attributable to maternal permanent environmental effects:

The total heritability:

The genetic correlation between the direct and maternal effects:

The direct-maternal correlation (r_am) was computed as the ratio of the estimates of the direct maternal covariance (σ_am) to the product of the square roots of the estimates of σa2 and σm2. The accuracy of the different model estimates was tested using the AIC as follow:

Where L represents the maximum likelihood value of the model, and k denotes the total number of parameters to be estimated.

A lower AIC value indicates a better balance between model fit and complexity, reflecting superior explanatory power while controlling for overfitting [20].

The BIC is a model selection criterion similar to AIC, balancing goodness-of-fit and complexity. The BIC was as follow:

Where L is the maximum likelihood value of the model, k is the number of parameters in the model, and n is the sample size. Consequently, the model with the smallest BIC among candidate models is typically selected as optimal.

To identify the most appropriate model for the target traits, a LRT was employed. This test evaluates the statistical significance of additional parameters by comparing the likelihood functions of nested models, ultimately determining the simplest model that retains strong explanatory power. A LRT was performed to obtain the appropriate model as given below [21]:

Where L₁ and L₂ denote the maximum likelihood values of model 1 and model 2, respectively. Model 1 is a sub-model of model 2. LR follows a chi-square distribution, with degrees of freedom equal to the difference in the number of parameters between Model 2 and Model 1. A statistically significant test result (p<0.05) indicates that the additional parameter(s) in Model 2 have a meaningful impact on the trait; otherwise, their effect is deemed non-significant. Furthermore, genetic and phenotypic correlations between traits were estimated using bivariate animal models. These models were constructed based on the optimal model selected for each trait. The structure of the bivariate model is as follows [22]:

Where y₁ and y₂ represent the vectors of observed traits 1 and 2. b₁ and b₂ are the vectors of fixed effects. a₁ and a₂ denote the vectors of additive genetic effects. e₁ and e₂ correspond the vectors of residual effects. The design matrices X₁ and X₂ associate observations with fixed effects, while Z₁ and Z₂ link the traits to their respective additive genetic effects. Genetic and phenotypic correlations between the two traits were estimated following the methodology outlined by Alam et al [23].

The description statistics for early growth traits are shown in Table 1. The average BW was 3.94 kg with a coefficient of variation (CV) of 18.38%. The average WW was 29.12 kg with a CV of 15.14%. The average ADG was 0.28 kg/d with a CV of 17.44%. The average KR was 0.02 with a CV of 6.69%. These values highlight significant trait variability, which is essential for the genetic assessment and improvement of Dorper sheep.

The least square mean and standard deviation for each trait of Dorper sheep are shown in Table 2. The effects of recipient dam age, sex, birth year, birth month and herd were highly significant (p<0.01) on early growth traits. In terms of recipient dam age, lambs born to 2-year-old and 4-year-old dams exhibited significantly higher BW than those born to 3-year-old dams (p<0.05). Conversely, offspring of 3-year-old dams achieved the highest WW, ADG, and KR, which were significantly greater than those of lambs from 2-year-old and 4-year-old dams (p<0.05). With respect to sex, male lambs outperformed females across all growth traits, with statistically significant differences observed for each trait (p<0.05). Regarding birth year effects, lambs born in 2019 and 2021 had significantly higher BW compared to those born in 2020 (p<0.05). For WW, ADG, and KR, the highest values were recorded in 2020-born individuals, while the lowest were observed in 2019-born lambs, with significant differences among the three years (p<0.05). Among birth months, lambs born in April displayed the highest BW (4.58±0.79 kg), which was significantly greater than that of lambs born in most other months (p<0.05). Lambs born in December achieved the highest WW (33.08±4.10 kg) and ADG (0.32±0.05 kg/d), whereas those born in July showed the lowest performance across all traits, with significant variations between different months (p<0.05). Comparisons across herds revealed that Herds 1, 5, and 9 had higher BW, with Herd 4 showing the lowest BW (3.41±0.68 kg). Herds 1, 2, and 5 performed better in terms of WW and ADG, while Herd 7 exhibited the lowest values. Significant differences were observed for most traits between herds (p<0.05).

The variance components and genetic parameters of BW, WW, ADG, and KR estimated by the six models are listed in Table 3. The heritability of BW estimated by different models ranged from 0.0215 (model 5) to 0.0484 (model 4). The maternal heritability ranged from 3.58×10⁻⁷ (model 5) to 0.0134 (model 4), and the variance ratio of the maternal permanent environmental effects was 0.2880 (model 2) to 0.2941 (model 6). The direct genetic correlation with maternal heritability was −0.9997 for model 4 and −0.9839 for model 6, showing a negatively correlated.

It was estimated that the correlation of WW heritability is between 0.0026 (model 6) and 0.0444 (model 4), maternal heritability is between 2.91×10⁻⁸ (model 5) and 0.0248 (model 4), the variance ratio of the maternal permanent environmental effects is be-tween 0.3185 (model 6) and 0.3214 (model 2), and that of direct genetic effect with ma-ternal heritability is −0.4121 (model 4) and −0.2054 (model 6).

The ADG heritability estimated by different models was between 0.0391 (model 2) and 0.1295 (model 6), the maternal heritability range was between 8.48×10⁻⁶ (model 5) and 0.0505 (model 4), and the variance ratio of the maternal permanent environmental effects was between 0.3090 (model 2) and 0.3219 (model 6). There was a highly negative correlation between the direct additive effect and maternal genetic correlation, which was estimated to be −0.8347 by model 6 and −0.7925 by model 6.

The KR heritability estimated by different models was between 0.0501 (model 5) and 0.1489 (model 6), the maternal heritability range was between 0.0003 (model 5) and 0.0726 (model 4), and the variance ratio of the maternal permanent environmental effects was between 0.2730 (model 5) and 0.2980 (model 6). The direct additive effect had a highly negative correlation with the maternal genetic effect, ranging from −0.9785 (model 6) to approximately −0.9556 (model 4).

The heritability, phenotypic and genetic correlation estimates between pre-weaning growth traits of Dorper sheep are represented in Table 6. The heritability of BW, WW, ADG, and KR were estimated to be 0.0215, 0.0287, 0.0391 and 0.0504, respectively, using optimal model (model 2) for each trait, which belonged to low heritability. BW was negatively genetically correlated with WW (−0.7546), ADG (−0.9728), and KR (−0.4991) traits. WW had a positive genetic correlation with ADG (0.9952) and KR (0.9984). WW also had a positive phenotypic correlation with ADG (0.9829) and KR (0.8819). Consistent with the negative genetic correlations between BW and WW/ADG/KR (Table 6), selecting for larger BW alone would hinder improvements in post-natal growth, emphasizing the need to prioritize maternal environmental factors (e.g., recipient milk quality). So indirect selection for ADG and KR could be achieved by selecting for WW, which gained its own faster genetic progression. WW can be used as an efficient selection index for combined genetic improvement of multiple traits.

The phenotypic characteristics of growth traits in sheep are shaped by a combination of genetic and environmental factors, including birth year, month, sex, herd, and maternal age. In Dorper sheep, recipient dam age exerted a highly significant effect (p<0.01) on BW, WW, ADG, and KR. This suggests that maternal physiological conditions during gestation and lactation directly influence early postnatal growth traits [25]. Herd effects significantly impacted all pre-weaning traits (p<0.01), highlighting the critical role of standardized management across breeding units. Variations in feeding regimens, health protocols, and housing conditions drove phenotypic divergence, a pattern consistent with findings in Chinese superfine merino sheep [26]. Temporal factors such as birth year and month also played pivotal roles, reflecting seasonal fluctuations in climate and management strategies. Similar temporal effects were observed in Mandya and Corriedale sheep, where birth year strongly influenced BW and WW [27,28]. Sex differences were pronounced in Dorper lambs, with males outperforming females across all traits, aligning with hormonal and metabolic ad-vantages reported in Avikalin sheep [29]. This contrasts with Sonadi sheep, where neither sex nor birth type affected KR [30], suggesting genetic or environmental inter-actions may mitigate sex-based growth disparities in certain breeds. Additionally, while Magotra [31] identified sex and birth type as critical drivers of BW, WW, and ADG across breeds, KR was solely influenced by sex, further emphasizing trait-specific sensitivity to environmental factors.

On the basis of environmental effects, we performed genetic parameters and compared six different animal models. The LRT revealed significant differences among the six animal models for all investigated traits (BW, WW, ADG, KR), with Model 2—incorporating both direct additive genetic effects and maternal permanent environmental effects—consistently emerging as the optimal model. This model demonstrated superior goodness of fit, evidenced by the lowest AIC values of traits (AIC = 130.4091 for BW, 7,056.4486 for WW, −10,644.0815 for ADG, and −24,327.6934 for KR). The maternal permanent environmental effects (e.g., uterine conditions, postnatal care) reduces the ha2 estimate, and they explain 28.8%, 32.14%, 30.9% and 27.31% of the phenotypic variation, respectively, aligning with findings in Dumeng sheep [4], where maternal environmental contributions were critical for pre-weaning trait. The results of Kushwaha et al [32] demonstrated that direct heritability estimates for all traits were significantly inflated when maternal permanent environmental effects were omitted from the model. Comparative model analyses revealed that biologically reliable and precise heritability estimates could only be achieved by simultaneously incorporating both direct additive genetic effects and maternal permanent environmental effects into the analytical model. Model 2’s optimal fit and low trait heritability are non-contradictory, reflecting the embryo transfer population’s characteristics. Low direct heritability hd2=0.0026-0.1489; Table 3) and total heritability ( ht2=0.0109-0.0565; Table 3) were primarily driven by two interrelated factors: The LOPU-IVF-ET technology decoupled the genetic source (donor ewes) from the environmental source (recipient ewes), as recipients exclusively determined fetal uterine development and postnatal lactation supply (Materials and Methods). This decoupling substantially diluted the contribution of direct additive genetic effects to phenotypic variation. Maternal additive genetic effects of donors were weakly expressed (maximum maternal heritability hm2=0.0726; Table 3), and negative genetic correlations between direct and maternal additive effects (r_am = −0.2054 to −0.9997; Table 3) further offset total genetic variance. Collectively, Model 2 provides the most accurate genetic parameters for guiding Dorper sheep breeding under LOPU-IVF-ET systems, with key implications: optimizing recipient ewe management (e.g., lactation nutrition regulation) will be critical to improving the efficiency of pre-weaning trait improvement.

The heritability estimates for pre-weaning traits in this study were consistently low, with clear population-specific and technical drivers. The heritability estimates for BW in this study (0.0215) was also similar to findings in Awassi sheep [33], was low heritability. Notably, while Tesema et al [34] reported higher total heritability (0.10 for BW) in Dorper × indigenous sheep, this discrepancy primarily arises from population genetic background and reproductive management differences: their crossbred population inherently harbors greater additive genetic variation due to heterosis, whereas our purebred Dorper population (14 sires, 328 dams; Table 1) has limited genetic diversity. Additionally, Dorper×indigenous sheep relied on natural breeding, where maternal genetic and environmental effects are biologically linked, whereas our use of LOPU-IVF-ET decouples the genetic source (donor ewes) from the environmental source (recipient ewes). In our study, fetal development and postnatal lactation depend entirely on the recipient’s uterine environment and milk supply (Materials and Methods), which dilutes direct additive genetic effects—this explains why our BW heritability (0.0215) is lower than that of South African Dorper sheep (0.07) raised under natural breeding [16]. WW exhibited low heritability (0.0287) in this study, in line with trends in Dumeng sheep [4], but lower than estimates for Alpine Merino sheep [35] and Jordan Awassi sheep [36], possibly due to the smaller population size here (n<3,000 vs. n>5,000 in other studies), as larger populations generally yield more accurate heritability estimates. ADG also exhibited low heritability (0.0391) in this study, lower than that reported for Central Anatolian Merino sheep [37], Mecheri sheep [38], Boer goats [39], and Barbari goats [40]. KR showed low heritability (h² = 0.0504), comparable to Sangsari sheep [14], models that included direct additive genetic effects and maternal permanent environment effects had significant effects on KR. Large values for maternal permanent environmental effects consistent with the biological characteristics of ruminant embryonic development under assisted reproduction. Maternal permanent environmental effects strongly impacted early growth traits, as lambs depend on maternal milk supply pre-weaning and are sensitive to environmental changes (e.g., feeding conditions). Small changes in the feeding environment, nutritional quality of feed, drinking water conditions and feeding management level of the ewes may have a direct or indirect significant effect on the WW of the lambs, preweaning daily weight gain and other growth indicators. Thus, the effect of maternal permanent environment is particularly prominent in this critical period before weaning.

Estimates using the best animal model 2 showed that the absolute values of the genetic correlation coefficients for BW, WW, ADG, and KR ranged from 0.4991 to 0.9984. BW was negatively genetically correlated with WW (−0.7546), ADG (−0.9728), and KR (−0.4991) traits. The study of Behrem [37] stated that the genetic correlation coefficient between newborn weight and WW of Merino sheep was −0.30, and the study of Arzik et al [11] concluded that the genetic correlation between BW and WW was negative, which is the same as the results of the present study. Negative genetic correlations between BW and post-weaning growth traits suggesting that direct selection for BW may have an antagonistic effect on the improvement of later growth traits to avoid compromising overall breeding progress by simply pursuing initial BW. WW had a positive genetic correlation with ADG (0.9952) and KR (0.9984). WW also had a positive phenotypic correlation with ADG (0.9829) and KR (0.8819). The high genetic correlation between ADG, WW and KR is similar to that expressed by Similar results to Ghafouri-Kesbi et al [41] estimate of 0.91 for Mehraban sheep and Gholizadeh and Ghafouri-Kesbi [42] estimate of 0.93 for Baluchi sheep, and Shokrollahi and Baneh [43] in Arabi sheep. So indirect selection for ADG and KR could be achieved by selecting for WW, which gained its own faster genetic progression. WW can be used as an efficient selection index for combined genetic improvement of multiple traits.