Optimizing Breeding Strategies for Pekin Ducks Using Genomic Selection: Genetic Parameter Evaluation and Selection Progress Analysis in Reproductive Traits

Abstract

Reproductive performance is an important trait in poultry production. Traditional methods of improving reproductive traits can only use recorded information from females, making it difficult to effectively assess the reproductive potential of males. Although genomic selection is thought to remedy this shortcoming, most studies now use simulated data or one or two generations of data to assess its effects. Also, the effectiveness of genomic selection for use in the improvement of reproductive traits in ducks has hardly been reported. In this study, data from four consecutive generations of Pekin duck populations were used to assess the effect of genomic selection on reproductive trait improvement. Whole-genome resequencing was performed for genotyping, and pedigree and SNP genetic parameters were evaluated. Using the BLUP (Best Linear Unbiased Prediction), GBLUP (Genomic Best Linear Unbiased Prediction), and ssGBLUP (Single-step Genomic Best Linear Unbiased Prediction) models, we assessed selection progress for body weight at 6 weeks, age at first egg, and egg number from 25 to 44 weeks over multiple generations. Ten-fold cross-validation was used to evaluate the genomic prediction performance. The results indicated that the heritability of growth traits decreased after routine selection, while reproductive and egg quality traits maintained moderate heritability (0.2–0.4). Selection progress showed a one-day advancement in age at first egg and an increase of one egg per generation from the 13th to 15th generations. The GBLUP model performance significantly outperformed BLUP, but ssGBLUP showed minimal improvement due to comprehensive genotyping. In conclusion, this study provides crucial insights for optimizing breeding strategies and improving economic efficiency in Pekin duck breeding.

1. Introduction

In poultry production, enhancing reproductive performance has always been a key objective. However, selective breeding for reproductive traits faces several obstacles. Firstly, reproductive traits are typically complex polygenic traits influenced by multiple genes and environmental factors [1]. The genetic mechanisms underlying these traits are complex and difficult to elucidate, making effective improvement by traditional phenotypic selection methods challenging [2]. The collection of phenotypic data is costly and time-consuming, requiring prolonged monitoring and recording and involving significant labor and resource investment. Traditional pedigree and phenotypic selection methods have limited performance in improving reproductive traits, with slow selection response and low precision. The main reason for this is that the reproductive performance of individuals of opposing sexes can almost only be replaced by records of full siblings.

Genomic selection, utilizing whole-genome information, enhances selection efficiency and accuracy, overcoming the limitations of traditional methods [3,4,5]. Genomic selection methods, such as GBLUP (Genomic Best Linear Unbiased Prediction), ssGBLUP (Single-step Genomic Best Linear Unbiased Prediction), and Bayesian approaches, have significantly increased breeding precision and efficiency [6,7,8]. Genomic breeding value estimation allows an accurate assessment of the genetic potential of individuals of opposite sexes, improving the efficiency of breeding plans. Genomic selection can be used to improve reproductive performance in agricultural animals such as rainbow trout [9], sheep [10], and dairy cattle [11]. However, widespread application still faces several challenges, including the complexity of genomic data and the diversity of breeding objectives. By integrating advanced genomic selection technologies with traditional breeding methods and leveraging big data and precision breeding tools, it is possible to gradually overcome these challenges and achieve the effective improvement of reproductive traits [12,13]. For meat ducks, as for other meat birds, selection indices focus on growth rates and feed conversion efficiency. Selection also continues for meat quality or fat traits to meet market needs. Selection for egg laying traits is carried out to further improve breeding efficiency. Most selection for these traits is carried out by phenotypic or sibling selection, with genomic selection becoming more common.

The Pekin duck is famous for its tender meat and high-quality eggs, occupies an important position in the poultry industry, and is a world-famous standard breed. As the consumption of duck meat and eggs increases year by year, the need for high-quality domestic duck breeds is becoming more and more urgent. Compared to pigs, chickens, and cows, the application of genomic selection technology to ducks is only just getting started. In our previous study, we found that genomic selection methods have high accuracy in the assessment of duck meat yield and growth traits [6]. However, research on genomic prediction in ducks remains limited and has mainly focused on meat quality traits [14], carcass traits [15], and growth and breast morphological traits [6]. The application of genomic selection techniques in duck reproductive traits has not yet been thoroughly explored.

Therefore, this study aimed to estimate the genetic parameters of reproductive traits (age at first egg and number of eggs between 25 and 44 weeks) and egg quality traits (average egg weight at 40 weeks and egg shape index) in multiple consecutive generations of Pekin ducks. Additionally, we evaluated the selection progress of age at first egg and number of eggs between 25 and 44 weeks over several generations using BLUP, GBLUP, and ssGBLUP models. We also assessed the genomic prediction performance of different models through 10-fold cross-validation. By comparing these data with relevant studies on other poultry species, we aim to further elucidate the genetic progress of these traits in Pekin ducks and the potential application of genomic selection. This will provide robust theoretical support and practical guidance for future poultry genetic improvement.

2. Materials and Methods

2.1. Population and Phenotype Measurement

The VII line Pekin duck population in this study was continuously selected for reproductive traits for more than 10 generations. Individual data (6045 ducks) from generations 12 to 15 were selected for analysis in this study. All individuals had detailed pedigree records, including 1132 males and 4913 females. For all individuals, we measured body weight (BW), body slope length (BSL, measuring the distance from the shoulder joint to the ipsilateral ischial tuberosity along the surface of the live bird using a tape measure), and neck length (NL, measuring the straight-line distance from the base of the neck to the end of the neck using a tape measure) at 6 weeks of age. For all female ducks, we accurately recorded the age at first egg (AFE) and detailed egg production records from 25 to 44 weeks of age (EN25-44WK). Additionally, at 40 weeks of age, we collected three eggs from each individual and measured their average egg weight (EWT) and average egg shape index (ESI, the ratio of the longitudinal diameter to the transverse diameter of the egg).

The experimental period commenced when the ducks were six weeks old and concluded at forty-four weeks of age. During this period, ducks were raised free-range from six to seventeen weeks of age and housed individually in cages from eighteen to forty-four weeks of age, with one male duck assigned per four female ducks for mating. Regarding light management, natural daylight with an intensity of 5–10 lux was provided from six to seventeen weeks of age. From eighteen to forty-four weeks of age, the photoperiod was adjusted according to production requirements, while the light intensity was maintained at 20 lux. In terms of feed formulation, from six to seventeen weeks of age, the diet provided 2700 kcal/kg of metabolizable energy (ME), 15.5% crude protein (CP), 1.0% lysine, and 0.3% methionine. From eighteen to forty-four weeks of age, the diet was formulated to supply 2800 kcal/kg ME, 20.5% CP, 1.1% lysine, and 0.45% methionine. The duck housing utilized a closed shed design with windows to ensure adequate ventilation and a suitable living environment. Across different generations and batches, all rearing management conditions were kept consistent to ensure the comparability and reliability of the experimental data.

2.2. Genotyping and SNP Identification

Blood samples were drawn from the metatarsal vein using standard venipuncture and collected in vacuum tubes. Genomic DNA was extracted using a QIAampR DNA Blood Mini Kit (QIAGEN, Tian Gen Biotech (Beijing) Co., Ltd., China). Whole-genome resequencing was performed on the DNBSEQ-T7 platform with 150 bp paired-end reads, achieving an average depth of over 2× per generation. For subsequent analysis, reads were aligned to the mallard duck reference genome (ASM874695v1) [16] using BWA (v0.7.10) [17]. Following alignment, SNPs were identified using GATK HaplotypeCaller (v4.1) [18], with parameters set to default except for -stand_call_conf being set to 30. Subsequently, the individual gvcf files were merged, and autosomal quality control was performed using VCFtools (v0.1.16) [19] with the parameters --remove-indels, --minQ 30, --min-alleles 2, and --max-alleles 2 to filter SNP sites for genotype imputation. After this, STITCH (v1.6.10) [20] was used for genotype imputation. Post-imputation, quality control was conducted using BCFtools (v1.8) [21] with INFO_SCORE > 0.4. Then, PLINK (v 1.90) [22] was used for further quality control based on --maf 0.01, --geno 0.05, and --mind 0.05. Following this, linkage disequilibrium pruning was performed with the parameters 50 5 0.2 to extract independent SNP sites. Finally, 365,860 SNPs were retained for subsequent analysis (Figure S1).

2.3. BLUP, GBLUP, and ssGBLUP Models

To estimate breeding values, we employed several methodologies utilizing different data types: BLUP for pedigree and phenotypic data [23], GBLUP for genotypic and phenotypic data [24], and ssGBLUP for a combination of pedigree, genotypic, and phenotypic data [25]. Our linear mixed model is expressed as follows:

$$y=X\mu +Zu+e.$$

(1)

In this equation, $y$ represents the vector of observations, and $\mu$ denotes the vector of fixed effects, which includes factors such as sex and batch. $X$ is the design matrix corresponding to the fixed effects, $u$ stands for the vector of additive genetic effects, $Z$ is an incidence matrix allocating records to $u$, and $e$ represents random residual effects.

For the BLUP approach, the vector of additive genetic effects $u$ is assumed to follow a normal distribution with a mean of zero and variance $A{\sigma }_a^2$, where A is the numerator relationship matrix derived from pedigree information, ${\sigma }_a^2$ is the additive genetic variance, and ${\sigma }_e^2$ is the residual variance. Heritability ($h^2$) is calculated as the ratio of additive genetic variance to the total phenotypic variance, given by

$$h^2={\displaystyle \frac{{\sigma }_a^2}{{\sigma }_a^2+{\sigma }_e^2}}.$$

(2)

Heritability is classified as high (>0.4), moderate (0.2–0.4), or low (<0.2) [26]. Correlation coefficients are categorized as negligible (0.0–0.2), weak (0.2–0.4), moderate (0.4–0.7), or strong (0.7–1.0) [27].

In the GBLUP method, the additive genetic effects vector $u$ is modeled with a normal distribution $N\left(0, G{\sigma }_a^2\right)$, where G is the genomic relationship matrix calculated using VanRaden’s method [28]. The genomic relationship matrix G is defined as follows:

$$G={\displaystyle \frac{Z{Z}^{\prime }}{2\sum \begin{array}{c}m\\ i=1\end{array}{p}_i\left(1-p_i\right)}},$$

(3)

where Z is the centralized genotype matrix, which can be calculated using the marker genotypic matrix M [24]. The genotype matrix M is an n × m dimensional matrix, where n is the sample size, m is the number of markers, and the elements in M are 0 (homozygote), 1 (heterozygote), or 2 (other homozygote); $p_i$ is the frequency of the minor allele of the i-th marker. The matrix $Z$ is defined as follows:

$$Z=M-2p_i.$$

(4)

For the ssGBLUP method, the additive genetic effects vector $u$ follows a normal distribution $N(0, H{\sigma }_a^2)$, where H is a blended relationship matrix combining both pedigree and genomic information. The inverse of H is derived using the following formula:

$$H^{-1}=A^{-1}+\left[\begin{array}{cc}0& 0\\ 0& G^{-1}-A_{22}^{-1}\end{array}\right],$$

(5)

where $A^{-1}$ is the inverse of the pedigree relationship matrix, $A_{22}^{-1}$ is the inverse of the numerator relationship matrix for genotyped animals, and $G^{-1}$ is the inverse of the genomic relationship matrix.

Variance components and estimated breeding values (EBVs) or genomic estimated breeding values (GEBVs) were determined using restricted maximum likelihood (REML) as implemented in the ASReml-R software (v4.2) [29].

2.4. Prediction Assessment

We performed a linear regression analysis for each phenotypic trait to extract the residuals adjusted for fixed effects (sex and batch) as the corrected phenotypes (for reproductive traits and egg quality traits in female ducks, the fixed effects include only batch effects). The definition of the corrected phenotype is as follows:

$${\widehat{\epsilon }}_i=P_i-({\widehat{\beta }}_0+{\widehat{\beta }}_1·Sex+{\widehat{\beta }}_2·Batch),$$

(6)

where ${\widehat{\epsilon }}_i$ is the residual, representing the corrected phenotype; $P_i$ is the observed phenotypic value; $Sex$ is the fixed effect of sex; $Batch$ is the fixed effect of batch; and ${\widehat{\beta }}_0$, ${\widehat{\beta }}_1$, ${\widehat{\beta }}_2$ are the regression coefficients estimated through linear regression.

To estimate the model’s predictive performance based on 10-fold cross-validation, we randomly divided the animals with both phenotypic and genotypic data into 10 subsets. In each iteration, the phenotypic values of one subset were set to “NA”, and the remaining 9 subsets were used as a reference group to predict the GEBVs of the subset. The accuracy of genomic prediction was defined as the Pearson correlation coefficient between the corrected phenotypic values and the GEBVs, expressed as

$$Accuracy=r={\displaystyle \frac{Cov\left(Y, \widehat{Y}\right)}{{\sigma }_Y{\sigma }_{\widehat{Y}}}},$$

(7)

where $r$ is the Pearson correlation coefficient; Y is the corrected phenotypic value; $\widehat{Y}$ is the GEBVs predicted by the model; $Cov\left(Y, \widehat{Y}\right)$ is the covariance between Y and $\widehat{Y}$; ${\sigma }_Y$ is the standard deviation of Y; and ${\sigma }_{\widehat{Y}}$ is the standard deviation of $\widehat{Y}$.

The unbiasedness of genomic prediction was defined as the regression coefficient between the GEBVs and the corrected phenotypic values, obtained from the linear regression model described as

$$Y={\beta }_0+{\beta }_1\widehat{Y}+\epsilon ,$$

(8)

where Y is the corrected phenotypic value; $\widehat{Y}$ is the GEBVs predicted by the model; ${\beta }_0$ is the intercept; ${\beta }_1$ is the regression coefficient; and $\epsilon$ is the error term. The measure of unbiasedness is that the regression coefficient ${\beta }_1$ should be close to 1, i.e.,

$$Unbiasedness={\beta }_1.$$

(9)

3. Results

3.1. Description of Phenotypic Data

In this study, we conducted a detailed analysis of the growth traits, reproductive traits, and egg quality traits of 6045 Pekin ducks over four consecutive generations. Regarding growth traits, the body weight at 6 weeks of age showed a declining trend. Specifically, in the 15th generation, the body weight at 6 weeks of age decreased by approximately 40 g compared to the 14th generation (

Table 1. Descriptive statistics of growth traits at 6 weeks of age for Pekin ducks.

Traits 1	Generation	Num 2	Mean	SD 3	CV (%) 4	Min	Max	Num (M) 5	Mean (M)	Num (F) 6	Mean (F)
BW (g)	12	709	2878	105	3.66	2682	3098	\	\	709	2878 (105)
	13	1638	2818	108	3.85	2630	3090	339	2970 (55)	1299	2778 (80)
	14	1878	2815	130	4.60	2555	3150	409	2968 (92)	1469	2772 (104)
	15	1820	2768	107	3.87	2565	3015	384	2923 (48)	1436	2726 (76)
BSL (cm)	12	708	25.31	0.87	3.44	22.9	27.2	\	\	708	25.31 (0.87)
	13	1629	25.59	0.92	3.61	22.7	28.19	331	26.28 (0.79)	1298	25.41 (0.87)
	14	1878	24.34	0.80	3.27	22.1	26.5	409	25.28 (0.58)	1469	24.07 (0.63)
	15	1817	24.38	0.76	3.11	22.5	26.5	384	25.38 (0.51)	1433	24.11 (0.57)
NL (cm)	12	709	20.25	0.65	3.19	18.1	21.8	\	\	709	20.25 (0.65)
	13	1626	20.39	0.95	4.66	17.34	23.3	331	21.57 (0.86)	1295	20.09 (0.71)
	14	1878	20.92	0.84	4.03	18.7	23.6	409	22.04 (0.66)	1469	20.61 (0.59)
	15	1819	20.74	0.82	3.95	18.8	23.1	384	21.85 (0.59)	1435	20.45 (0.59)

Note: ¹: BW, body weight at 6 weeks of age; BSL, body slope length; NL, neck length. ²: Number of ducks that pass quality control of phenotypic value. ³: Standard deviation. ⁴: Coefficient of variation. ⁵: M denotes male ducks. ⁶: F denotes female ducks.

, Figure S2). Similarly, the body slope length decreased by approximately 1 cm in the last two generations compared to the first two, while the neck length increased by about 0.5 cm (Table 1, Figure S2). In terms of reproductive traits, the age at first laying gradually advanced with each generation. The 15th generation started laying about one week earlier than the 14th generation (

Table 2. Descriptive statistics of reproductive traits for Pekin ducks.

Traits 1	Generation	Num 2	Mean	SD 3	CV (%) 4	Min	Max
AFE	12	704	176.04	11.57	6.57	148	211
	13	1217	174.20	12.26	7.04	146	210
	14	1395	173.51	11.22	6.47	147	206
	15	1362	165.89	14.09	8.49	134	206
EN25-44WK	12	691	119.55	11.96	10.00	73	140
	13	1199	120.03	12.38	10.32	76	140
	14	1378	120.53	12.31	10.22	77	141
	15	1342	124.05	11.91	9.60	78	140

Note: ¹: AFE, age at first egg; EN25-44WK, egg number from 25 to 44 weeks of age. ²: Number of ducks that pass quality control of phenotypic value. ³: Standard deviation. ⁴: Coefficient of variation.

, Figure S3). Additionally, the number of eggs laid between 25 and 44 weeks of age steadily increased with each generation. The 15th generation laid approximately three more eggs during this period compared to the 14th generation (Table 2, Figures S3 and S4). Regarding egg quality traits, the average egg shape index at 40 weeks of age remained relatively stable across generations (

Table 3. Descriptive statistics of egg quality traits at 40 weeks of age for Pekin ducks.

Traits 1	Generation	Num 2	Mean	SD 3	CV (%) 4	Min	Max
EWT	12	681	89.53	5.89	6.58	72.6	105.3
	13	1216	90.24	5.32	5.89	74.5	106.5
	14	1354	90.68	5.45	6.01	75.8	107
	15	1361	88.28	5.09	5.77	73.7	104
ESI	12	680	1.36	0.04	3.14	1.24	1.48
	13	1211	1.36	0.04	3.28	1.24	1.49
	14	1352	1.36	0.04	3.11	1.23	1.49
	15	1362	1.36	0.04	3.08	1.23	1.48

Note: ¹: EWT, average egg weight at 40 weeks of age; ESI, average egg shape index at 40 weeks of age. ²: Number of ducks that pass quality control of phenotypic value. ³: Standard deviation. ⁴: Coefficient of variation.

, Figure S3). However, the average egg weight at 40 weeks of age in the 15th generation decreased by 2.4 g compared to the 14th generation (Table 3, Figure S3).

3.2. Estimation of Heritability and Genetic Correlation

The genetic parameters for growth traits, reproductive traits, and egg quality traits are shown in

Table 4. Genetic parameters of growth, reproductive, and egg quality traits ¹.

Traits 2	BW	BSL	NL	AFE	EN25-44WK	EWT	ESI
BW	0.19 (0.02) 0.11 (0.02)	0.52 (0.05) **	0.12 (0.07)	0.11 (0.06)	–0.14 (0.07) *	0.19 (0.06) **	0.03 (0.06)
BSL	0.39 (0.09) **	0.27 (0.02) 0.20 (0.02)	0.39 (0.06) **	–0.01 (0.06)	0.06 (0.06)	0.24 (0.05) **	0.07 (0.06)
NL	0.27 (0.11) **	0.25 (0.09) **	0.23 (0.02) 0.16 (0.02)	–0.04 (0.06)	0.12 (0.06) *	0.09 (0.06)	–0.01 (0.06)
AFE	0 (0.1)	–0.22 (0.08) **	–0.08 (0.09)	0.37 (0.02) 0.26 (0.03)	–0.91 (0.02) **	0.06 (0.05)	–0.07 (0.05)
EN25-44WK	–0.04 (0.11)	0.22 (0.08) **	0.14 (0.09)	–0.89 (0.04) **	0.27 (0.02) 0.24 (0.03)	–0.03 (0.06)	0 (0.06)
EWT	0.32 (0.10) **	0.24 (0.08) **	0.08 (0.09)	–0.06 (0.08)	0.16 (0.09) *	0.39 (0.02) 0.24 (0.03)	–0.03 (0.05)
ESI	–0.01 (0.11)	0.14 (0.08) *	–0.09 (0.09)	–0.17 (0.08) *	0.03 (0.09)	0 (0.09)	0.36 (0.02) 0.23 (0.03)

Note: ¹: Heritability is given on the diagonal (bold for SNP-based heritability, italic bold for pedigree-based heritability), SNP-based genetic correlations are shown above the diagonal, and pedigree-based genetic correlations are shown below the diagonal. The standard errors of the estimates are provided in parentheses. In the correlation coefficients between two traits, “**” indicates p ≤ 0.01, and “*” signifies p ≤ 0.05. ²: BW, body weight at 6 weeks of age; BSL, body slope length; NL, neck length; AFE, age at first egg; EN25-44WK, egg number from 25 to 44 weeks of age; EWT, average egg weight at 40 weeks of age; ESI, average egg shape index at 40 weeks of age.

. The heritability estimates based on SNPs are generally higher than those based on pedigrees. In this study population, the pedigree-based results indicate that growth traits have low heritability. Similarly, the SNP-based results show that growth traits have low to moderate heritability. For reproductive traits and egg quality traits, both pedigree-based and SNP-based heritability estimates indicate that these traits have moderate heritability. This is related to our breeding practice of prioritizing selection for body weight, which results in lower heritability estimates for growth traits compared to reproductive traits. The genetic correlation results based on pedigrees and SNPs show that the three 6-week growth traits have weak correlations (0.2–0.4). According to the SNP data, there is a moderate correlation (0.52 ± 0.05) between BW and BSL. Both the pedigree-based and SNP-based results indicate a strong correlation (0.7–1) between AFE and EN25-44WK, while the correlations among egg quality traits are negligible. Additionally, there is a weak correlation between EWT, BW, and BSL.

3.3. Selection Response

We compared the average GEBVs and EBVs of BW, AFE, and EN25-44WK across successive generations for male ducks, female ducks, and the overall population. As shown in Figure 1 and Figure S5, the GEBVs and EBVs for BW in males, females, and the overall population showed a decreasing trend over generations, despite some fluctuations. For reproductive traits, the direction of selection was more apparent. The GEBVs and EBVs for the AFE in males, females, and the overall population decreased progressively across generations. From the 13th to the 15th generation, the AFE advanced by approximately one day per generation. Correspondingly, the GEBVs and EBVs for the EN25-44WK increased progressively across generations. From the 13th to the 15th generation, the egg number increased by one egg per generation. Compared to the EBVs, the range of the GEBV changes across successive generations was broader, indicating potentially higher predictive accuracy. In summary, these results suggest that genomic selection has the potential to improve breeding accuracy and is significant for optimizing the growth and reproductive traits of Pekin ducks.

3.4. Genomic Prediction Using BLUP, GBLUP, and ssGBLUP Models

Figure 2 presents the results of predicting seven traits using the pedigree BLUP, GBLUP, and ssGBLUP models. These results were obtained through 10-fold cross-validation, demonstrating the prediction accuracy and unbiasedness. The prediction accuracy based on the BLUP model was relatively low, ranging from 0.10 to 0.23. In contrast, the prediction accuracy based on the GBLUP model was moderate, ranging from 0.20 to 0.43. Compared to the BLUP model, the GBLUP model showed a significant improvement in prediction accuracy, with an average increase of 0.15 across the seven traits. Specifically, the average increases for growth traits, reproductive traits, and egg quality traits were 0.11, 0.14, and 0.21, respectively. Using SNP information, the prediction accuracy for the EWT improved the most, reaching an increase of 107.72%. The average prediction accuracy for the seven traits increased by 80.52%. The prediction unbiasedness of different models was close to 1. Compared to the BLUP model, the GBLUP model improved the overall prediction unbiasedness by 1.42% across the seven traits. Specifically, the prediction unbiasedness for growth traits, reproductive traits, and egg quality traits increased by an average of 0.09%, 2.26%, and 2.57%, respectively. The prediction unbiasedness for BSL decreased the most, by 1.97%, while EN25-44WK showed the greatest improvement, with an increase of 5.34%. When combining the pedigree and SNP information, the prediction accuracy and unbiasedness based on the ssGBLUP model were relatively consistent with the results obtained from the GBLUP model.

4. Discussion

4.1. Phenotypic and Genetic Parameters

The breeding population in this study underwent routine selection. Growth traits were measured at 6 weeks of age, and based on production requirements, individuals within a specific weight range were selected. Simultaneously, the top 5% of individuals with the longest BSL and NL were culled to enhance the uniformity of the breeding population. After routine selection, the coefficients of variation for growth phenotypic traits were relatively low (3.11–4.66%). In this study, multiple consecutive generations of Pekin duck populations were primarily focused on improving reproductive performance to enhance egg production efficiency. By implementing body size selection at six weeks of age, the uniformity of the population was increased. However, this approach also led to a decrease in the heritability of growth traits, such as body weight, body slope length, and neck length. The primary reasons for the decline in heritability include the reduction in phenotypic variation due to selective culling and the shift in selection pressure towards reproductive performance. Selective breeding may result in a decrease in genetic diversity, leading to underestimated heritability estimates for related traits [30]. Additionally, changes in selection pressure can affect the estimation of genetic variance [31]. In unselected chicken populations, growth traits tend to exhibit relatively high heritability [32]. Research showed that the heritability of growth rate in chicken populations remained above 0.4 even after challenging the birds with the La Sota strain of Newcastle Disease Virus, compared to a heritability of 0.55 prior to the challenge [33].

Based on pedigree and SNP information, the heritability estimates for reproductive traits and egg quality traits in this study population reached moderate levels (0.2–0.4). There was a strong negative correlation between AFE and EN25-44WK, but a weak correlation between EWT and ESI. Similar results have been observed in related studies on other poultry populations [8,34,35,36,37,38,39]. Early onset of laying is usually associated with higher reproductive efficiency because these individuals enter the peak laying period earlier [38]. A smaller age at first egg (AFE) indicates earlier sexual maturation in individuals [40], which is associated with certain key genetic variations or the regulation of endocrine axes [41]. During the peak laying period from the onset of laying to 44 weeks of age, breeding female ducks are in their highest egg production phase. A smaller AFE allows individuals to lay more eggs within the fixed laying cycle (25–44 weeks of age), thereby increasing the number of eggs produced during this period (EN25-44WK). Consequently, there is a strong negative correlation between AFE and EN25-44WK. This finding is consistent with our previous research and other related studies [34,42,43]. In addition, egg weight is mainly influenced by the volume and mass of albumen and yolk, whereas the shape index is the ratio of the egg’s length to its diameter. These traits have different genetic and physiological bases, resulting in a weak correlation between the egg weight and shape index [39].

In the studied population, a weak genetic correlation was observed between growth traits at 6 weeks of age and both egg quality traits at 40 weeks of age and egg production traits from 25 to 44 weeks of age. This weak correlation can be attributed to several factors. First, there is a significant discrepancy in the timing of trait measurements; growth traits are evaluated during early developmental stages, whereas egg quality and egg production traits are assessed at sexual maturity and later stages. This temporal mismatch diminishes the genetic association between these traits [34,44]. Second, the genetic architectures of these traits may be relatively independent, with distinct sets of genes governing growth traits and egg quality and egg production traits, thereby reducing their genetic correlation [45]. Additionally, gene–environment interactions may lead to varying genetic correlations for different traits under the environmental conditions corresponding to different time periods. In summary, the weak genetic correlation between growth traits at 6 weeks of age and egg quality traits at 40 weeks of age, as well as egg production traits from 25 to 44 weeks of age, results from the combined effects of multiple factors. Understanding and addressing these factors is essential for optimizing breeding strategies and achieving the coordinated development of growth traits alongside egg quality and egg production traits.

4.2. Improvements in the Selection Process

Research indicates that the heritability estimates for growth traits are generally high, allowing for genetic improvement through direct phenotypic selection [46]. After routine breeding in the studied population, the uniformity of growth traits improved. Despite fluctuations in the EBVs and GEBVs for BW over four consecutive generations, the overall trend was a decline. To optimize the growth performance and structure of the population, managers implemented early phenotypic selection to cull individuals with extreme body weights or disproportionate body and neck lengths, thereby enhancing the uniformity of the breeding population. This strategy not only improves population consistency but also ensures that breeding goals align better with market demands and production realities.

In terms of reproductive traits, particularly from the 13th to 15th generations, the EBVs and GEBVs for AFE decreased across generations, while those for EN25-44WK increased. Specifically, each generation’s AFE was on average one day earlier than the previous generation, and EN25-44WK increased by one egg. This study separately evaluated the genetic gains for AFE and EN25-44WK over multiple generations in different sexes and overall populations, using both pedigree and SNP information. The results showed that the selection progress for AFE and EN25-44WK in males was consistent with that in females from the 13th to 15th generations.

By integrating pedigree and SNP data, the genetic parameters can be estimated more accurately, thereby optimizing selection strategies for reproductive traits. Research has demonstrated that SNP-based selection can significantly enhance the accuracy and efficiency of genetic gain, particularly in large populations and long-term breeding programs [47,48]. Additionally, this method helps to better understand the role of genetic variation in reproductive traits and provides reliable data support for future breeding plans.

4.3. Genomic Selection Predictive Performance

Based on our study population, our findings indicate that compared to the traditional BLUP method, genomic prediction using SNP information significantly improves prediction accuracy while maintaining an unbiasedness close to 1. Specifically, for egg quality traits, the prediction accuracy doubled with the use of the GBLUP model. Other studies have similarly reported varying degrees of improvement in genomic selection prediction accuracy with the GBLUP model compared to the BLUP model [6,49,50]. The higher accuracy of breeding value predictions using SNP information may be due to the more precise estimation of genetic relationships compared to traditional pedigree methods [6]. In this study, the heritability estimates for growth traits decreased after conventional selection, resulting in lower prediction accuracy. In contrast, the heritability estimates for reproductive traits and egg quality traits were moderate, leading to relatively higher prediction accuracy. This suggests a positive correlation between heritability estimates and prediction accuracy [7]. Compared to the GBLUP method, the ssGBLUP method allows for a more comprehensive integration of genomic data and pedigree data, thereby improving the prediction accuracy of genomic selection [25]. Additionally, the ssGBLUP method is more efficient in handling large-scale data, particularly in the prediction of complex traits in poultry and livestock [24]. In our study, the prediction accuracy for various traits showed almost no improvement from the GBLUP to the ssGBLUP model. This may be because all 6045 individuals in the study population were genotyped, a finding that has also been corroborated by similar study [6].

Nevertheless, unlike BLUP, which constructs the relationship matrix exclusively using pedigree information, ssGBLUP integrates both pedigree and genomic data to develop the relationship matrix H. This integration more accurately reflects individual differences and significantly enhances the accuracy of genetic assessments. Consequently, ssGBLUP serves as an excellent method for combining traditional breeding data with genotyping information. Furthermore, utilizing traditional pedigree records in conjunction with partial genotyping can effectively reduce breeding costs while improving prediction accuracy. The data source population in this study comprises a specialized line selected for egg laying traits, with a breeding program aimed at maximizing egg laying performance while ensuring appropriate growth rates and feed conversion efficiency. Reproductive traits are critical for breeding efficiency; however, traditional selection methods have been ineffective in improving these traits. The results of this study demonstrated significant improvements in egg laying ability after more than three consecutive generations of selection using a genomic selection program. Compared to traditional selection methods, genomic selection offers superior accuracy for selecting egg laying traits in Pekin ducks. Additionally, the one-step method effectively integrates genealogical and genotypic information, thereby reducing breeding costs. These findings suggest that incorporating genomic selection and partial genotyping into breeding strategies can enhance trait improvement efficiency and economic viability in future breeding programs.

5. Conclusions

This study conducted a comprehensive genetic parameter estimation for the reproductive traits and egg quality traits of Pekin ducks and evaluated the selection progress and genomic prediction performance over multiple generations using BLUP, GBLUP, and ssGBLUP models. The selection progress analysis indicated that from the 13th to the 15th generation, the AFE advanced by one day per generation, and the EN25-44WK increased by one per generation. The GBLUP model significantly outperformed the BLUP model in prediction performance, but the ssGBLUP model showed minimal improvement due to all individuals being genotyped. By integrating genetic parameter estimation results and selection progress analysis, we confirmed the important role of genomic selection technology in optimizing breeding strategies for Pekin ducks. In the future, by leveraging higher-precision genomic data and effective selection strategies, breeders can further enhance the breeding efficiency and economic benefits of Pekin ducks, providing robust scientific evidence and practical guidance for genetic improvement in poultry.