Research Note: Analysis of body weight and walking ability in turkeys and the prediction of categorical responses across systematic effect classes using a linear threshold model

Abstract

Heavy selection for growth in turkeys has led to a decay in leg soundness and walking ability. In this study, different models and traits were used to investigate the genetic relationships between body weight (BW) and walking ability (WA) in a turkey population. The data consisted of BW and WA traits collected on 276,059 male birds. Body weight was measured at 12 and 20 wk and WA at 20 wk of age. For WA, birds were scored based on a 1 (bad) to 6 (good) grading system. Due to the small number of records with scores 5 and 6, birds with WA scores of 4, 5, and 6 were grouped together resulting in only 4 classes. Additionally, a binary classification of WA (scores 1 and 2 = Similarly, an estimate of the genetic correlation between WA and BW at 20 wk was −0.45, indicating a more pronounced class 1; scores 3, 4, 5, and 6 = class 2) was evaluated. The inheritability estimates of WA ranged between 0.25 and 0.27 depending on the number of classes. The Heritability of BW at 12 and 20 wk was 0.44 and 0.51, respectively. The genetic correlation between WA and BW at 12 wk was around −0.35, indicating that heavy birds tend to have poor WA. antagonistic relationship between BW and WA. The genetic correlation between BW at 12 and 20 wk was positive and high (0.80). The residual correlation between WA and BW at 12 and 20 wk of age was −0.07 and −0.02, respectively. The residual correlation between body weight traits was 0.57. Similar results were observed when a binary classification was adopted for WA. The probability of an individual with a given genetic merit expressing a certain class of WA was determined for different fixed effect designations. Predictive probabilities clearly showed that birds when hatched in the winter would have a small chance to exhibit good WA phenotypes.

INTRODUCTION

Selection for fast growth in turkeys resulted in reduced fitness. Heavy birds often suffer from leg weakness, joint problems, and bone fractures. Mobility-related traits are of crucial importance due to their direct association with productivity and bird welfare. Several traits including shank diameter, leg muscle, drum bones, and gait scores were used to assess mobility and leg soundness. These traits are under the influence of diverse genetic and environmental factors and some of them are highly correlated. Existing data point towards a negative relationship between growth and mobility in turkeys and several studies showed clear differences between selected and nonselected turkey birds (Havenstein et al., 1988; Hiscock et al., 2022).

For commercial turkeys, birds from breeding lines are subjectively scored for their walking ability (WA) using a 6-grade scoring system with 1 = poor and 6 = good. These subjective scores are derived based on 7 mobility-related features (motion, pitch, balance, leg angulation, hock strength, hip strength, and leg structure). A detailed description of the scoring system can be found in Quinton et al. (2011). Heritability estimates of leg disorders are generally low ranging between 0.01 and 0.15 (Kapell et al., 2017). For WA, heritability estimates using a linear model ranged between 0.18 and 0.25 (Quinton et al., 2011). Genetic correlation estimates between WA and body weight (BW) traits are moderate and negative. However, Kapell et al. (2017) reported positive correlations between BW and mobility-related traits.

Walking ability, as it is currently scored by the turkey industry, is a discrete trait with 6 classes derived based on 7 mobility-related features. The relatively large number of classifying factors (7) and number of classes for WA (6) are likely to increase the inconsistencies (misclassification) of the subjective assessment of the trait. Furthermore, WA is often analyzed as a continuous response in clear violation of its discrete distributional form. The objectives of this study are to investigate (1) the relationship between growth traits and WA, (2) the potential impact of a simplification of the walking scoring system using large commercial data and the appropriate statistical modeling approach, and 3) the effect of the genetic merit on the probabilities of predicting the different categorical responses for WA across different classes of systematic effects.

MATERIAL AND METHODS

Data: Growth and WA scores of turkey birds hatched between 2009 and 2018 were used. The phenotypic data consisted of records on 276,059 male birds and the pedigree included 836,785 animals. Birds were measured for BW at 12 (BW12) and 20 (BW20) wk of age, and WA at 20 wk of age using a 1 (bad) to 6 (good) scoring system. Due to the small number of observations with WA scores of 5 (n = 53) and 6 (n = 1), birds with these scores were joined with those with scores 4 (n = 1,580) making a unique class (score 4) with 1,634 records. To assess the impact of the simplification of the scoring system and to potentially reduce the uncertainties associated with the 6-class scoring system, a binary classification was used where birds with scores 1 and 2 were assigned to class 1, and the remaining birds were assigned to class 2.

Data analysis: A threshold-linear mixed model was implemented for the joint analysis of WA, WB12, and BW20. The following mixed linear model was assumed:

$$y_{ijk}=H_{ik}+{\beta }_k*{\left(age\right)}_{jk}+u_{jk}+e_{ijk}$$ (1)

where $y_{ijk}$ is the observed phenotype for BW12, and BW20 (k = 1,2), or the liabilities for WA (k = 3) for bird j, H_i was the fixed effect of the hatch wk class i,${\beta }_k$ is the regression coefficient of bird age for trait k (only for growth traits), ${\left(age\right)}_{jk}$is the age of bird j for trait k, $u_{jk}$ is the random additive effect of bird j for trait k, and $e_{ijk}$is the random residual term.

A full Bayesian analysis was carried out to implement the model presented in equation [1] following the methodology and software developed by Rekaya et al. (2013).

Predictive probabilities: Under the assumed model, the liability for a bird i follows a normal distribution $l_i\sim N\left({\eta }_i,{\sigma }_e^2\right)$ with mean ${\eta }_i=x_i\beta +u_i$. The vector $\beta \text{and}{u}_i$ are the vector of fixed effects and the breeding value, respectively, and ${\mathit{x}}_i$ is a vector linking fixed effects to the liability.

Under the assumed threshold model, the relationship between the observed ordered categorical response and the liabilities is given by:

$$c_i=jif{t}_{j-1}<l_i<t_j\text{for}j=1,2,\dots ,4$$ (2)

where $c_i=j$ is the categorical response for animal i is equal class $j\left(j=1,2,\dots ,4\right),\text{and}{t}_k\text{is}\text{the}\text{threshold}k\left(k=0,1,\dots ,4\right)$with $t_0=-\infty$ and $t_4=+\infty$

Given the assumed distribution for the liabilities and the relationship in equation [2], the probability of observing class $j\left(j=1,2,\dots ,4\right)$ for animal i is given by:

$$pr\left(c_i=j|\beta ,u_i,{\sigma }_e^2\right)=\Phi \left(t_j|{\eta }_i,{\sigma }_e^2\right)-\Phi \left(t_{j-1}|{\eta }_i,{\sigma }_e^2\right)$$

where $\Phi \left(t_j|{\eta }_i,{\sigma }_e^2\right)$ is the cumulative distribution function of a normal distribution with mean equal to ${\eta }_i$ and variance equal to ${\sigma }_e^2$ evaluated at $t_j$.

Season effects (Winter = December–February; Spring = March–May; Summer = June–August; and Fall = September–November) were approximated by adjusting the hatching week estimates for the year effects.

RESULTS AND DISCUSSION

For both WB12 and WB20, there was a decreasing trend with the increase in WA scores. The differences between the mean BW12 were small with a maximum of around 120 g (∼1.5%) between birds in classes 1 and 4. For BW20, the differences were larger with the maximum difference of 584 g (∼4%). Differences in BW at 12 and 20 wk of age for birds in adjacent WA classes were negligible. However, there was substantial variation within and across WA classes as indicated by the relatively large standard deviations of around 1.0 and 1.4 for BW12 and BW20, respectively. When WA was classified as binary, similar distribution for the three traits was observed both in trend and magnitude of the differences. For both classifications, a certain level of phenotypic association between WA and BW emerges, and that increases with age. However, the differences point to substantial overlap in BW across WA classes.

Estimates of heritability, genetic, and residual correlations for the three traits using a linear-threshold model are presented in Table 1. Estimates of heritability for growth traits were moderate to high ranging between 0.44 and 0.51 independently of the classification of WA. These estimates are similar to those reported by (Rafat et al., 2011), but they are substantially larger than the estimates reported by Quinton et al. (2011) also using male lines. The major difference between the study of Quinton et al. (2011) and the current study was that their statistical model assumed WA as a continuous trait. For WA, the point estimates of heritability ranged between 0.24 and 0.27. These estimates are similar to those obtained by Kestin et al. (2001) and Quinton et al. (2011). However, they are significantly higher than the 0.06 reported by Havenstein et al. (1988). On top of the marked differences in methodologies and models used, the definition of the trait itself may have not been consistent across these studies. Walking ability, like other discrete traits, has in general a low heritability. However, environmental and management factors play a crucial role in the expression of these traits (liabilities exceeding a given threshold to express the observed trait). In a breeding population with optimum environmental and management conditions, a greater genotypic expression is expected and consequently a higher heritability. This is well supported with the results presented in Figure 1. The genetic and residual correlations between WA using the 4 classes scoring system and growth traits were moderate (−0.45 to −0.35) and weak (−0.07 to −0.02), respectively (Table 1). The genetic correlation between the 2 growth traits was, as expected, high (0.80); however, the residual correlation was moderate (0.57). Almost identical results were observed when WA was classified as a binary trait (Table 1).

Table 1.

Estimates of heritability (diagonal), genetic (above diagonal), and residual (below diagonal) correlations between body weight traits walking ability defined as 4 classes and binary response (SD are between parentheses).

	4 classes			Binary response
	WA	BW 12	BW 20	WA	BW 12	BW 20
WA	0.25	−0.35	−0.45	0.27	−0.34	−0.44
	(0.01)	(0.03)	(0.02)	(0.02)	(0.02)	(0.02)
BW 12	−0.07	0.51	0.80	−0.10	0.50	0.80
	(0.01)	(0.03)	(0.04)	(0.01)	(0.03)	(0.04)
BW 20	−0.02	0.57	0.44	−0.07	0.57	0.44
	(0.01)	(0.02)	(0.04)	(0.01)	(0.03)	(0.03)

Abbreviations: BW12 = body weight at 12 wk of age; BW20 = body weight at 20 wk of age; WA = walking ability at 20 wk of age.

Figure 1.

Predictive probabilities of the different categorical responses for walking ability across different systematic classes using the average breeding values in (A) class 1 (−0.156), (B) class 2 (0.0829), (C) class 3 (0.299), and (D) class 4 (0.532).

These results clearly indicate that, at least at the genetic level, selection for heavier birds at 12 or 20 wk of age will lead to WA problems. Furthermore, the problem tends to accentuate with the age of the bird. These results seem to indicate that fast-growing birds will allocate a large portion of nutrients to accrue muscle and fat deposition at the expense of skeleton development. Even when enough nutritional resources are available the problem persists due to the unbalance between the rate of growth and skeleton maturity and the biological limitations on the amount of feed a bird can consume. Several studies have shown similar results. In fact, Ye et al. (1997) reported a negative phenotypic association (−0.34) between breast yield and WA in selected male turkeys at 16 wk of age. Nestor et al. (2005) reported a negative genetic correlation (−0.29) between BW and WA at 20 wk of age using male turkey birds. Quinton et al (2011) reported similar results to the findings of the current study for the genetic correlations between WA using the 6 classes scoring system and BWs at 15 and 20 wk of age. However, few studies (Leishman et al., 2023) reported moderate positive genetic relationships between growth traits and WA.

When WA was classified as binary, there were marginal changes in the heritability and correlations compared to using the 4-class scoring system. The heritability of WA increased by 0.02 (0.27 vs. 0.25). For BW traits, heritability decreased by 0.01 for BW12 and remained the same for BW20. No noticeable change was observed in the estimates for the genetic and residual correlations (Table 1). These results seem to support the binary classification of WA removing the potential inconsistencies associated with a 6-class scoring system.

Figure 1 presents the predictive probabilities of a bird with genetic merit equal to the average breeding values (BVs) of animals in each of the WA classes having an observed phenotype for WA of 1, 2, 3, or 4 across three different classes of fixed effects (bad, average, and good) calculated based on the estimates of the fixed effects used in the model. Estimates for season effects were −0.034, 0.001, 0.012, and −0.001 for Winter, Spring, Summer, and Fall, respectively. Based on those estimates, fixed effects classified as “bad” was winter, “average” was Spring and Fall, and “good” was Summer. For a bird with genetic merit equal to −0.156 (average BVs for birds in WA class 1), its probability of having a WA score phenotype of 1 ranged between 71% when hatched in winter (bad environment) to 21% when hatched in summer (good environment) as indicated in Figure 1A. The same bird will have almost a zero probability of having a WA score phenotype of 4, independently of the season within which it is hatched (0.01–4.5%). For a bird with genetic merit equal to 0.532 (average BVs for birds in WA class 4), it will have 5%, 17, and 38% probability for a WA phenotype of 1 when hatched in summer, spring/fall, and winter, respectively. The same bird will have a probability ranging between 1.4% and 20% that its phenotype for WA is 4 across the different seasons (Figure 1D) with the highest probability when hatched in the winter. When the bird has genetic merit equal to 0.083 and 0.229 (average BVs for birds in WA classes 2 and 3), intermediate results were observed with a steady decrease in the probability of observing a WA phenotype score of 1 and an increase in the probability of observing a WA phenotype score of 4 with the increase in the average merit. This alludes to a significant genotype by season effect even for a population under a standardized environmental, nutritional, and management controls.

Our results showed a substantial variation in growth traits within and across WA classes indicating the possibility for additional genetic improvement of growth traits without, at least, further deterioration in WA. Using a 6-class system to score birds for WA might not be the most appropriate. Our findings showed that a more simplified scoring system (e.g., 4 classes) or even a binary classification could be sufficient. To improve WA, hatching season may play an important role for birds with average genetic merit for WA to express the desired mobility phenotypes as clearly shown by the predictive probabilities.