Analysis of Social Genetic Effects on Pigs Fed With Automatic Feeders Using a Visit‐Based Approach

ABSTRACT

Social behaviour traits and their impact on feed efficiency are of particular interest in pig farming. The integration of automatic feeders enables the collection of multiple phenotypes for breeding purposes. The additive genetic and social genetic effect can be estimated considering all the visits to the feeder by modelling each visit independently in a ‘visit‐based approach’. This study aimed to determine the impact of the social genetic effect on individual feed intake and duration per visit in Pietrain pigs and Iberian pigs separately. The dataset comprised 883,906 visits from 1608 Pietrain pigs and 775,054 visits from 856 Iberian pigs. In the Pietrain population, the social genetic effects did not explain a substantial percentage of the phenotypic variance (~1%). In contrast, the Iberian population exhibited more substantial contributions, with social genetic effects accounting for 6.2% of the variance in duration per visit and 5.5% in feed intake per visit. The correlations between additive direct genetic and additive social genetic effects were slightly positive for feed intake across all analyses, and around zero for duration per visit with most of them including the zero in the highest posterior density interval (HPD95%). These weak correlations suggest that both effects could be selected independently. The visit‐based approach successfully identified social genetic effects in the studied populations. Models incorporating social genetic effects demonstrated lower residual variance, enhancing the accuracy of additive values and, consequently, the potential for an improved response to selection.

1. Introduction

The incorporation of automatic feeders in pig farms has been steadily increasing. This technology facilitates the recording of multiple phenotypes for management and breeding purposes. Several studies have estimated the genetic parameters of feed efficiency and feeding behaviour traits in ‘white pigs’ fed by automatic feeders (Chen et al. 2010; de Haer, Luiting, and Aarts 1993; Do et al. 2013; Kavlak and Uimari 2019; Labroue et al. 1999; Núñez et al. 2023; Santiago et al. 2021). The data produced by automatic feeders can also be used to examine social interactions among animals, a factor that can impact feed efficiency traits (Piles et al. 2024). Bijma, Muir, and Van Arendonk (2007) proposed a model that incorporates these interactions, including direct and indirect genetic effects. In this model, each individual interacts with $n–1$ of its group members, where $n$ is the group size. The overall breeding value of an individual $i$ is equal to ${\mathrm{TBV}}_{\mathrm{i}}={\mathrm{d}}_{\mathrm{i}}+\left(\mathrm{n}-1\right){\mathrm{s}}_{\mathrm{i}}$, where $\mathrm{d}$ and $\mathrm{s}$ denote the additive direct genetic and additive social genetic effects, respectively. In this model, the social genetic effect is an associative effect that can be interpreted as a heritable environmental effect provided by the pen mates to the focal individual (Wolf et al. 1998), being equivalent to a model with maternal‐effects. This model has been employed to estimate genetic parameters of feed efficiency (Bergsma et al. 2008; Bijma, Muir, and Van Arendonk 2007) and feeding behaviour traits (Herrera‐Cáceres, Ragab, and Sánchez 2019; Kavlak et al. 2021) in pig populations. In the above‐mentioned studies, the phenotype consisted of a single value per animal, usually the average of all visits per individual.

An alternative approach was proposed by Angarita et al. (2021), where each visit was individually modelled. This proposal makes use of all the information generated by automatic feeders. In this proposal, the additive genetic effect is attributed to the pig eating at each visit, while the social genetic effect belongs to the pig entering the feeder immediately after (‘follower‐pig’). It is important to note that this model estimates the social genetic effect considering all the possible interactions between animals in the same group. However, Angarita et al. (2021) only focused on the duration per visit, and they only modelled the data at the phenotypic level due to the lack of pedigree information.

The aim of this study was to assess the influence of the social genetic effect on feed intake and duration per visit in Pietrain and Iberian pig populations by using all the visits recorded by automatic feeders. We based this study on the preliminary work of Angarita et al. (2021), but we expanded it by incorporating pedigree information and adding feed intake per visit as an additional trait. In addition, this study provided valuable information about genetic parameters of feeding behaviour traits in Iberian pigs under an intensive system, allowing to narrow the knowledge gap in comparison with the studies focused on ‘white’ pig populations.

2. Materials and Methods

2.1. Databases and Filtering Description

The databases used in this study belonged to two distinct breeds of pigs that were fed using automatic feeders (Nedap, Groenlo, the Netherlands). The first database corresponded to a purebred Pietrain male line from the selection nucleus of the company Selección Batallé S.A. (Riudarenes, Catalunya, Spain). The animals entered the testing period with an initial weight of 51.4 ± 8.4 kg and an age of 95.0 ± 7.5 days and finished at 113.2 ± 9.8 kg and an age of 162.4 ± 6.7 days. The Pietrain pigs were housed in groups of 12.7 ± 1.8 animals per pen (average density of 1.5 m² per pig) and were provided ad libitum access to a commercial diet. Data collection by automatic feeders spanned from June 2019 to October 2022, recording 883,906 visits from 1608 Pietrain male pigs across 699 litters. The pedigree consisted of 5217 individual‐sire‐dam entries from 5 generations. The median of the genetic relationship between pigs was 0.06. Additional details regarding diet, housing conditions, and the automatic feeder system are described in Núñez et al. (2023), where the same database was used to estimate genetic parameters of feeding behaviour and feed efficiency traits.

The second database comprised 775,054 recorded visits from 856 Iberian male pigs across 487 litters. The animals were housed in the Testing Center of the company IngaFood S.A. (Almendralejo, Extremadura, Spain). The pigs belonged to Entrepelado (EE) and Retinto (RR) strains, as well as their reciprocal crosses (ER and RE), where the first letter indicates the sire line and the second the maternal line of the cross. One or two animals per litter were included in the experiment in 80% of the total litters. The animals were randomly mixed and raised in pens with an average of 14.1 pigs per pen (with a mean pen density of 1.1 m² per animal). The maximum number of litter mates sharing the same pen was two. The data were collected from November 2020 to May 2023, and only visits up to 180 days of age were considered. The animals entered the testing period with an initial weight of 33.1 ± 10.4 kg and an age of 111.6 ± 24.5 days and finished with 77.8 ± 13.0 kg at 180 days of age. The pedigree information included 1957 individual‐sire‐dam entries containing the 856 tested animals and their ancestors, with a maximum of 5 generations. The median of the genetic relationship between pigs was 0.07.

The databases were initially filtered, eliminating visits lacking animal ID and those with zero feed intake and duration. The visits of the first 10 days were removed, as they were considered an adaptation period for pigs to the feeder machine. Subsequently, the datasets were filtered to remove anomalous values in feed intake and duration per visit. After an exploratory analysis, a threshold of 4 standard deviations (SD) was employed to identify outliers following the criteria described in Casey, Stern, and Dekkers (2005) (see Figures S1 and S2). The feed intake thresholds were established at 754 g for the Pietrain breed and 765 g for the Iberian breed, while the duration per visit thresholds were set at 1185 s for Pietrain and 1211 s for Iberian pigs. Owing to the similarity observed in the distributions, these thresholds were unified. Consequently, all the visits with feed intake > 760 g, duration > 1200 s or feeding rate > 298 g/min were excluded from the datasets. We conducted principal component analysis using the three filtered variables to visually assess the presence of outliers before and after filtering.

After data preprocessing, the filtered database contained 728,769 visits of 1608 pigs in the Pietrain database. The average number of visits per pig was 452 ± 139 visits and the number of days with records in automatic feeders was 68 ± 7.8 days. For Iberian pigs, the filtered database contained 749,345 visits from 856 pigs with an average of 874 ± 470 visits per pig. The number of days with records in Iberian pigs was 61 ± 19 days, which was more heterogeneous than in Pietrain. This variability in Iberian pigs was mainly due to a higher variability in age at the start of the study (111.6 ± 24.5 days) and, to a lesser extent, to non‐recorded cases due to ear tag losses and management practices (e.g., veterinary treatments) resulting in animals leaving the original pen at different times. However, ear tag losses and veterinary interventions did not affect a large number of animals (31 out of 856). All the animals had more than 20 days of records, while 80% of the animals had more than 40 days of records.

The visits were organised by location (feeder machine number) and visit time to incorporate the information about the ‘follower’ pig. The ‘follower’ pig was defined as the animal entering the feeder machine immediately after the current visit. Thus, each row in the database represented an individual visit, with its respective animal ID, feed intake per visit, duration per visit, body weight, visit time, follower ID and the interval between visits. As the objective of the study was to evaluate the effect of the ‘follower’ pig on the visit of the preceding pig, all visits where the animal ID and follower ID matched were excluded, as they represent animals reentering into the feeder machine.

The complete set of visits in each database (728,769 visits for Pietrain and 749,345 visits for Iberian) were classified as competitive and non‐competitive according to the criteria established in Angarita et al. (2021). Consecutive visits with intervals of < 60 s were designated as competitive, while those with intervals exceeding 600 s (10 min) were categorised as non‐competitive. The purpose of this classification was to examine the additive social effect at both extremes separately, expecting it to be more pronounced in competitive visits. Moreover, the analysis was also performed with the full database without classification.

2.2. Statistical Models

The genetic parameters for feed intake and duration per visit were estimated through univariate Bayesian analyses using two distinct models. Model 1 (Equation 1) was the standard animal model:

$$\mathbf{y}=\mathbf{Xb}+{\mathbf{Z}}_{\mathbf{d}}{\mathbf{a}}_{\mathbf{d}}+{\mathbf{Z}}_{\mathbf{p}}\mathbf{p}+\mathbf{e}$$ (1)

being y a vector of observations for each visit; X an incidence matrix relating the phenotypic data with the systematic effects b, which included the pen (animals sharing the same feeder machine), the hour of the day (see Figure S3), the strain for Iberian pigs, and also a linear regression coefficient for the days of age at a specific visit (covariate); ${\mathbf{Z}}_{\mathbf{d}}$ corresponds to an incidence matrix of random additive direct genetic effects ${\mathbf{a}}_{\mathbf{d}}$; ${\mathbf{Z}}_{\mathbf{p}}$ is the incidence matrix of animal permanent environmental effects $\mathbf{p}$, and $\mathbf{e}$ is a random vector of residual effects. Normal distributions were assumed for ${\mathbf{a}}_{\mathbf{d}}$, $\mathbf{p}$ and $\mathbf{e}$ effects; ${\mathbf{a}}_{\mathbf{d}}\mid {\mathrm{\sigma }}_{\mathrm{a}}^2~\mathrm{N}\left(0,\mathbf{A}{\mathrm{\sigma }}_{\mathrm{a}}^2\right)$, $\mathbf{p}\mid {\mathrm{\sigma }}_{\mathrm{pe}}^2~\mathrm{N}\left(0,\mathbf{I}{\mathrm{\sigma }}_{\mathrm{pe}}^2\right)$, respectively, where $\mathbf{A}$ is the additive relationship matrix, $\mathrm{I}$ is the identity matrix with the same order of the number of records, and ${\mathrm{\sigma }}_{\mathrm{a}}^2$, ${\mathrm{\sigma }}_{\mathrm{pe}}^2$ and ${\mathrm{\sigma }}_{\mathrm{e}}^2$ are variances for the additive, permanent and residual effects.

Model 2 is the social‐genetic model (Equation 2) which includes the additive social‐genetic effects to consider the genetic influence of the follower pig. This model assumes that if the social genetic effect exists, it would be explained by an effect of the follower pig on the current pig. Hence, the ‘social model’ was:

$$\mathbf{y}=\mathbf{Xb}+{\mathbf{Z}}_{\mathbf{d}}{\mathbf{a}}_{\mathbf{d}}+{\mathbf{Z}}_{\mathbf{s}}{\mathbf{a}}_{\mathbf{s}}+{\mathbf{Z}}_{\mathbf{p}}\mathbf{p}+\mathbf{e}$$ (2)

where $\mathbf{b}$ is the same as in (Equation 1), ${\mathbf{Z}}_{\mathbf{d}}$, ${\mathbf{Z}}_{\mathbf{s}}$ correspond to known incidence for additive direct genetic and additive social genetic effects, respectively, where ${\mathbf{a}}_{\mathbf{d}}$ and ${\mathbf{a}}_{\mathbf{s}}$ are random vectors of the additive direct and the additive social genetic effects, with

$$\left[\begin{array}{c}{\mathbf{a}}_{\mathbf{d}}\\ {\mathbf{a}}_{\mathbf{s}}\end{array}\right]\mid {\mathbf{G}}_{\mathbf{0}} ~ \mathrm{MVN}\left(0,{\mathbf{G}}_{\mathbf{0}}\otimes \mathbf{A}\right)$$

in which

$${\mathbf{G}}_{\mathbf{0}}=\left[\begin{array}{cc}{\mathrm{\sigma }}_{\mathrm{a}}^2& {\mathrm{\sigma }}_{\mathrm{as}}\\ {\mathrm{\sigma }}_{\mathrm{as}}& {\mathrm{\sigma }}_{\mathrm{s}}^2\end{array}\right]$$

where ${\mathrm{\sigma }}_{\mathrm{s}}^2$ is the social additive genetic variance and ${\mathrm{\sigma }}_{\mathrm{as}}$ represents the genetic additive‐social covariance. The rest of the components of Model 2 (Equation 2) were the same as those of the additive genetic model (Equation 1). In Model 2, direct heritability (${\mathrm{h}}_{\mathrm{d}}^2$) was obtained as ${\mathrm{h}}_{\mathrm{d}}^2={\mathrm{\sigma }}_{\mathrm{a}}^2/{\mathrm{\sigma }}_{\mathrm{p}}^2$, and the total heritable variation (${\mathrm{t}}^2$) was obtained as ${\mathrm{t}}^2=\left({\mathrm{\sigma }}_{\mathrm{a}}^2+{\mathrm{\sigma }}_{\mathrm{s}}^2\right)/{\mathrm{\sigma }}_{\mathrm{p}}^2$, where${\mathrm{\sigma }}_{\mathrm{p}}^2$ is the phenotypic variance, obtained as ${\mathrm{\sigma }}_{\mathrm{p}}^2={\mathrm{\sigma }}_{\mathrm{a}}^2+{\mathrm{\sigma }}_{\mathrm{s}}^2+2{\mathrm{\sigma }}_{\mathrm{as}}+ {\mathrm{\sigma }}_{\mathrm{pe}}^2+{\mathrm{\sigma }}_{\mathrm{e}}^2$. The proportion of the phenotypic variance explained by the additive social genetic effect was obtained as ${\mathrm{s}}^2={\mathrm{\sigma }}_{\mathrm{s}}^2/{\mathrm{\sigma }}_{\mathrm{p}}^2$.

The software gibbsf90+ and postgibbsf90 from the blupf90 family (Misztal et al. 2002) were used to solve the different mixed model approaches. For that, flat priors were considered for systematic effects and co(variances) in all the models. A single chain of 1,000,000 samples was used to obtain the posterior marginal distributions of each unknown parameter, and the first 50,000 iterations were discarded as ‘burn‐in’. One sample was taken every 100 iterations to calculate statistic parameters from the posterior marginal distribution to avoid autocorrelation between the samples. Convergence was checked separately for each estimation with Geweke's test (1992) and by visual observation. The deviance information criterion (DIC) (Spiegelhalter et al. 2002) was used as a measure of the goodness of fit of the different models.

To gain a deeper understanding of how animals with high and low social additive effects influence the phenotype of their pen mates, we conducted an exploratory analysis of the relationship between social additive estimates and the phenotypic values of their predecessors.

3. Results and Discussion

3.1. Phenotypic Description

The phenotypic characterisation of feeding behaviour and productive traits of the Pietrain and Iberian populations was summarised in Table 1. Notable differences in final body weight (FBW), average daily gain (ADG) and feed conversion rate (FCR) were observed between the two pig populations. As expected, the Pietrain population exhibited greater ADG and lower FCR than Iberian pigs (Table 1). The mean for ADG obtained in Pietrain pigs was 905 g/day, and for FCR was 2.08 kg/kg. These findings align with previously reported values for this breed in other studies (Cámara et al. 2016; Edwards, Tempelman, and Bates 2006; Lenoir et al. 2022; Saintilan et al. 2013). In Iberian pigs, the mean ADG and FCR were 621 g/day and 3.66 kg/kg, respectively. The ADG value was in line with those reported for Iberian pigs, which ranged between 450 and 800 g/day (Barea, Nieto, and Aguilera 2007; Nieto et al. 2019; Serra et al. 1998). The variation in the latter cited studies primarily stems from differences in the age of entry into testing, with older animals exhibiting higher ADG. It is more challenging to discuss FCR, due to its limited reporting in Iberian pig studies. However, we found higher values of DFI in Iberian (2.29 kg/day) than those for Pietrain pigs (1.87 kg/day). Similar differences were reported by Morales et al. (2002), showing a DFI of 3.66 kg/day for Iberian and 2.63 kg/day for Landrace pigs raised in similar conditions until reaching 108 kg of FBW. In Núñez et al. (2023) was described a positive correlation between DFI and FCR (0.62). This may suggest that Iberian pigs are less efficient animals due to their higher values for DFI. Note that the disparities with the values reported by Morales et al. (2002) could be generated from differences in FBW between studies (108 vs. 77 kg).

TABLE 1.

Phenotypic means and standard deviations for feeding behaviour (FBT) and feed efficiency traits (FET) of Pietrain and Iberian populations.

Trait	Type	Unit	Pietrain (n = 1608)		Iberian (n = 856)
			Mean	SD	Mean	SD
Duration per visit	FBT	Seconds	415	296	315	339
Feed intake per visit	FBT	Grams	252	189	150	173
Visits per day (NVD)	FBT	Number	7.55	2.01	13.7	5.68
Final body weight (FBW)	FET	kg	113	9.78	76.7	13.0
Final age	FET	Days	162	6.77	172	17.0
Daily feed intake (DFI)	FET	kg	1.87	0.21	2.29	0.50
Average daily gain (ADG)	FET	Grams	905	102	621	81.4
Feed conversion rate (FCR)	FET	kg/kg	2.08	0.20	3.66	0.68

Regarding the feeding behaviour traits, the most important difference between populations was the number of visits per day, with a mean of 7.5 visits/day in Pietrain and 13.7 visits/day in Iberian pigs. A broad range of results were reported in the literature for ‘white’ pigs. For instance, Do et al. (2013) reported a mean of 8.8 visits/day in Landrace, 11.1 visits/day in Duroc and 18.1 visits/day in Yorkshire pigs, indicating notable differences among pig breeds. Labroue et al. (1999) specifically studied the Pietrain breed and reported means of 7.5 visits/day and 300 g/visit of feed intake, which were comparable to the results presented in this study. Unfortunately, there is a lack of information on feeding behaviour traits in Iberian pigs, limiting direct comparisons with the traits measured in this study. The feed intake per visit and the duration per visit were higher in Pietrain compared with Iberian pigs (Table 1), but it was offset by the higher number of visits observed in Iberian pigs.

3.2. Classification of Visits

The classification of visits is presented in Table 2. A higher percentage of visits were categorised as competitive in the Iberian database (80%) compared with the Pietrain database (45%). In Angarita et al. (2021), the proportion of competitive and non‐competitive visits was 78% and 8% in commercial crossbred pigs with a mean density between 1 and 1.4 m² per pig. In our study, the mean densities per pig were 1.5 m²/pig for Pietrain and 1.1 m²/pig for Iberian pigs. The Pietrain population started the test with a weight of 52 kg and finished at 180 days with 113 kg, while Iberian pigs began with 33 kg and finished at 180 days with 78 kg. Thus, the adjusted densities as m² by total kg of body weight on the pen were similar between both breeds (see Table 3).

TABLE 2.

Classification of visits in Pietrain and Iberian populations.

Trait	Unit	Pietrain (n = 1,608)		Iberian (n = 856)
		Value	%	Value	%
Total database	Visits	728,769	100	749,345	100
Competitive visits (< 60 s)	Visits	324,642	45	603,446	80
Non‐competitive visits (> 600 s)	Visits	126,821	17	33,015	4

TABLE 3.

Mean density indicators for Pietrain and Iberian populations.

Indicator	Unit	Pietrain (n = 1608)	Iberian (n = 856)
		Value	Value
Animals per pen	Pigs	12.7	12.6
Density per animal	m2/pig	1.50	1.12
Initial mean density (per kg BW)	m2/kg	0.02	0.02
Final mean density (per kg BW)	m2/kg	0.01	0.01

3.3. Variance Component Estimation

The posterior means of variance components, heritabilities (${\mathrm{h}}^2$), and ratios of social (${\mathrm{s}}^2$) and total genetic variance (${\mathrm{t}}^2$) for the duration and feed intake per visit in both populations and each subset were estimated with Model 1 (Equation 1) and Model 2 (Equation 2). In Model 1 (additive genetic model), the genetic effects were attributed only to the additive genetic effects. In Model 2 (social genetic model), the genetic effects were divided into additive direct genetic and additive social genetic effects. The estimates of heritability and ratios for social and permanent effect variances are shown in Tables 4 and 5 (the estimates of variance components are in Tables S1–S4). Note that our models do not include the litter or contemporary group as random effects. Due to the relatively small dataset, it is difficult to fit a model with many random effects. Thus, we tried to simplify the model as much as possible without affecting the results. To do so, we previously analysed the results by including different random effects in the model (see Tables S5 and S6). The incorporation of litter or contemporary group as random effects did not improve the goodness of fit. We also confirmed that the model we used did not considerably affect the estimates of social genetic effects.

TABLE 4.

Genetic parameters of duration per visit for Pietrain and Iberian populations with the 95% highest posterior density interval (in brackets).

Trait	Type	Analysis	Pietrain				Iberian
			$${\mathrm{h}}^2$$	$${\mathrm{s}}^2$$	$${\mathrm{p}}^2$$	$\mathrm{\rho }$ a,s	$${\mathrm{h}}^2$$	$${\mathrm{s}}^2$$	$${\mathrm{p}}^2$$	$\mathrm{\rho }$ a,s
Duration per visit	Comp	Model 1	0.13 [0.10, 0.16]	—	0.05 [0.04, 0.07]	—	0.07 [0.04, 0.11]	—	0.08 [0.05, 0.10]	—
	Comp	Model 2	0.13 [0.10, 0.16]	0.04 [0.03, 0.04]	0.05 [0.03, 0.07]	0.05 [−0.20, 0.12]	0.06 [0.03, 0.09]	0.08 [0.07, 0.09]	0.06 [0.04, 0.08]	−0.05 [−0.18 0.07]
	NC	Model 1	0.14 [0.10, 0.17]	—	0.07 [0.05, 0.10]	—	0.04 [0.02, 0.06]	—	0.08 [0.06, 0.10]	—
	NC	Model 2	0.14 [0.10, 0.17]	0.00 [0.00, 0.00]	0.07 [0.05, 0.10]	−0.31 [−0.71, 0.04]	0.04 [0.02, 0.07]	0.00 [0.00, 0.00]	0.08 [0.05, 0.10]	0.10 [−0.35, 0.54]
	Total	Model 1	0.12 [0.09, 0.15]	—	0.05 [0.03, 0.07]	—	0.06 [0.03, 0.09]	—	0.08 [0.06, 0.10]	—
	Total	Model 2	0.12 [0.09, 0.15]	0.01 [0.01, 0.01]	0.05 [0.03, 0.07]	0.03 [−0.04, 0.11]	0.06 [0.03, 0.08]	0.06 [0.06, 0.07]	0.06 [0.04, 0.08]	0.00 [−0.13, 0.11]

Abbreviations: ${\mathrm{h}}^2$ = classical heritability; ${\mathrm{p}}^2$ = total genetic variance; ${\mathrm{s}}^2$ = proportion of phenotypic variance explained by the social effect; Comp = competitive visits; Model 1 = additive model; Model 2 = social genetic model; NC = non‐competitive visits; $\mathrm{\rho }$ _a,s = correlation between additive genetic effect and social genetic effect.

TABLE 5.

Genetic parameters of feed intake per visit for Pietrain and Iberian populations with the 95% highest posterior density interval (in brackets).

Trait	Type	Analysis	Pietrain				Iberian
			$${\mathrm{h}}^2$$	$${\mathrm{s}}^2$$	$${\mathrm{p}}^2$$	$\mathrm{\rho }$ a,s	$${\mathrm{h}}^2$$	$${\mathrm{s}}^2$$	$${\mathrm{p}}^2$$	$\mathrm{\rho }$ a,s
Feed intake per visit	Comp	Model 1	0.07 [0.05, 0.10]	—	0.05 [0.03, 0.06]	—	0.11 [0.06, 0.16]	—	0.07 [0.03, 0.10]	—
	Comp	Model 2	0.08 [0.05, 0.10]	0.03 [0.03, 0.04]	0.05 [0.03, 0.06]	0.12 [0.03, 0.20]	0.08 [0.05, 0.11]	0.07 [0.06, 0.08]	0.06 [0.03, 0.08]	0.23 [0.12, 0.35]
	NC	Model 1	0.07 [0.05, 0.10]	—	0.06 [0.04, 0.07]	—	0.05 [0.02, 0.09]	—	0.10 [0.07, 0.12]	—
	NC	Model 2	0.08 [0.05, 0.10]	0.00 [0.00, 0.00]	0.05 [0.04, 0.07]	0.12 [−0.24, 0.44]	0.06 [0.02, 0.10]	0.00 [0.00, 0.00]	0.09 [0.05, 0.11]	0.22 [−0.13, 0.56]
	Total	Model 1	0.07 [0.05, 0.09]	—	0.04 [0.02, 0.05]	—	0.10 [0.06, 0.14]	—	0.07 [0.03, 0.10]	—
	Total	Model 2	0.07 [0.05,0.09]	0.01 [0.01, 0.01]	0.04 [0.02, 0.05]	0.08 [−0.00, 0.16]	0.08 [0.04, 0.11]	0.05 [0.04, 0.06]	0.06 [0.03, 0.08]	0.25 [0.14, 0.37]

Abbreviations: ${\mathrm{h}}^2$ = classical heritability; ${\mathrm{s}}^2$ = proportion of phenotypic variance explained by the social effect; ${\mathrm{t}}^2$ = total genetic variance; Comp = competitive visits; Model 1 = additive model; Model 2 = social genetic model; NC = non‐competitive visits; $\mathrm{\rho }$ _a,s = correlation between additive genetic effect and social genetic effect.

In Pietrain pigs, the heritability estimates with both the additive genetic (Equation 1) and social effect model (Equation 2) were the same when the total database was analysed, showing values of 0.12 for duration per visit and 0.07 for feed intake per visit (Tables 4 and 5). This consistency was also observed in the competitive and non‐competitive visits. Moreover, low values were identified for the additive social genetic effect, as evidenced by the percentage of the total phenotypic variance explained for the additive social genetic effect (${\mathrm{s}}^2$) that were 3% for competitive visits, 1% for the total database, and null for non‐competitive visits. Thus, we concluded that the additive social genetic effect is not relevant for Pietrain pigs. This aligns with the findings reported by Angarita et al. (2021).

Conversely, the Iberian population showed more interesting outcomes. First, in the competitive visits, the estimate of ${\mathrm{h}}^2$ for duration per visit was 0.07 in the additive genetic model and 0.06 in the social genetic model (Table 4). For the feed intake per visit, the estimate of ${\mathrm{h}}^2$ was 0.11 for the additive genetic model and 0.08 for the social genetic model (Table 5). This result indicates that a little portion of the additive genetic variance estimated in Model 1 was attributed to additive social genetic variance when this effect was incorporated into the Model 2. The additive social genetic effect explained 8% of the total phenotypic variance for the duration per visit and 7% for feed intake per visit. Furthermore, the estimate of total heritable variation (${\mathrm{t}}^2$) with Model 2 (0.14) was higher than the ${\mathrm{h}}^2$ obtained with Model 1 (0.07). The estimate of the additive social genetic effects in Model 2 seems to come from the residual variance of Model 1, since no significant changes in the additive variance were found. This indicates that the inclusion of the additive social genetic effect could improve the model fit by reducing the residual variance, which may improve the selection response.

In the total database, the social genetic effect explained 6% of the total phenotypic variance for duration per visit, and 5% for feed intake per visit. Likewise, the estimate of ${\mathrm{t}}^2$ of Model 2 (0.12) was higher than ${\mathrm{h}}^2$ of Model 1 (0.06). It is important to emphasise that the incorporation of the social genetic effects (Equation 2) increased the total heritable variance, reducing the residual variance in the presence of competition between contemporaneous animals. Furthermore, it could potentially enhance the accuracy of the breeding values estimations and improve the response to selection for feed intake in Iberian pigs.

Minimal impact was noted for the additive social genetic effect in the non‐competitive visits for both traits, yielding no relevant differences between the ${\mathrm{h}}^2$ of the additive genetic model with the ${\mathrm{t}}^2$ of the social genetic model. It should be noted that non‐competitive visits entail longer intervals (> 600 s). Therefore, no significant social effects were expected in this subgroup, aligning with the results presented in this study (Table 2). Further information on the DIC (Spiegelhalter et al. 2002) for all the analyses performed is shown in Table 6. The DIC did not change in Pietrain population when additive social genetic effects were included. However, in Iberian pigs the DIC was lower when the additive social genetic effects were added for the analysis on the total data base and competitive visits, indicating a better fit of the Model 2 compared with Model 1.

TABLE 6.

Deviance information criterion (DIC) for analyses performed in Pietrain and Iberian populations for Model 1 and Model 2 and the traits duration and feed intake.

Trait	Model	Comp	Pietrain (n = 1,608)		Iberian (n = 856)
			NC	Total	Comp	NC	Total
Duration per visit	1	4,764,212	1,778,859	10,463,865	8,818,760	493,911	10,973,883
	2	4,756,008	1,778,846	10,457,664	8,786,432	493,890	10,943,870
Feed intake per visit	1	4,442,289	1,671,366	9,830,450	8,003,795	448,543	9,955,269
	2	4,434,860	1,671,352	9,825,507	7,978,230	448,506	9,915,498

Abbreviations: Comp = competitive visits; Model 1 = additive model; Model 2 = social genetic model; NC = non‐competitive visits.

No previous studies investigating social genetic effects in Iberian pigs fed with automatic feeders have been found. Iberian pigs are recognised for their high adaptation to extensive systems (Temple et al. 2011). Consequently, the degree of adaptation to intensive systems for Iberian pigs may be comparatively lower than that of ‘white pigs’, making this breed particularly interesting to study social effects. This could potentially explain our results, indicating why is easier to find competition, i.e., social genetic effect, when Iberian pigs are housed in limited space. Nevertheless, this study had different conditions between both populations, such as the type of management, diet and housing conditions. These factors make it challenging to accurately compare both populations. For this reason, this study was focused on describing the results for each population independently. It is important to highlight that this study was based on the study done by Angarita et al. (2021), but we incorporated the pedigree information to account for additive direct genetic and additive social genetic effects. Incorporating the genealogical information is useful to consider the relatedness between animals, providing more accurate estimates of genetic parameters.

3.4. Additive Direct and Additive Social Genetic Correlation

The correlations between additive direct genetic and additive social genetic effects were slightly positive for feed intake across all analyses, and around zero for duration per visit with most of them including the zero in the highest posterior density interval (HPD95%) (Tables 4 and 5). Positive correlations indicate that animals with beneficial additive social genetic effects also possess favourable additive direct genetic effects on the trait. On the other hand, the weak correlation suggests that both effects could be selected independently. One exception was found in the non‐competitive visits for the Pietrain population, where the correlation between additive direct genetic and additive social genetic effects was negative for duration per visit (−0.31). However, this estimation had a wide HPD95%, as it was obtained from only 17% of the total database. The literature exhibits a broad spectrum of values for the correlation between additive and social effects. Herrera‐Cáceres, Ragab, and Sánchez (2019) found correlations between direct and social genetic effects of −0.78 ± 0.27 for occupation time and −0.57 ± 0.47 for daily consumption in Duroc pigs. On the other hand, Kavlak et al. (2021) investigated the same correlation in Finnish Yorkshire for different periods. They reported positive correlations for the duration per visit. Additionally, they reported a decrease from 0.57 ± 0.24 in the first period to 0.30 ± 0.20 in the last period of study. In the feed intake per visit, the correlation remained close to 0.25 ± 0.21 throughout the five periods. In both studies (Herrera‐Cáceres, Ragab, and Sánchez 2019; Kavlak et al. 2021), the traits were an average of n visits, where the additive social genetic effects are considered as the associative effect provided from pen mates to the focal individual. This social effect was obtained considering that all the pen mates interact with the focal animal. In contrast, our study employed a visit‐based approach, as proposed by Angarita et al. (2021), where the social genetic effect was determined by considering each combination of feeder‐pig and follower‐pig, modelling all visits independently. It is important to clarify that in this approach, the social genetic effect (high or low) of an animal (follower‐pig) influences the phenotype value of the previous animal (feeder‐pig). This approach is advantageous because individuals do not necessarily have the same level of interaction, and the frequency of interactions between different pairs can vary. Consequently, the visit‐based approach is a valuable approximation for estimating additive social genetic effects.

3.5. Relationship Between Additive Social Genetic Effect and the Phenotype of Pen Mates

Both additive direct genetic and additive social genetic effects were estimated considering all the ‘feeder‐follower’ interactions. We computed the Pearson correlation between the additive social genetic value of the ‘follower’ pig and the mean feed intake of its predecessors (Figure 1). We observed a correlation of 0.71 and coefficient of determination (R ²) of 0.51, meaning that pigs with higher social breeding values, i.e., pigs less competitive, contribute to an increase in feed intake of the animal in the feeder. Specifically, when considering only the top 10% of animals in the ranking for additive social effect in feed intake, the mean feed intake at the previous visits was 238.1 ± 57.4. In contrast, the previous visits of the bottom 10% had a mean of 120.8 ± 28.9. Hence, pigs with negative additive social genetic values have a negative effect on the feed intake of the animals in the feeder, which may indicate a highly competitive behaviour. However, confirming this ‘behavioural theory’ is not possible, as no behaviour observations were recorded in this study.

FIGURE 1.

Relationship between the social breeding value of the follower pig and the phenotypic mean of feed intake of the preceding pigs. [Colour figure can be viewed at wileyonlinelibrary.com]

4. Conclusions

The visit‐based approach is a simple and valuable approximation for assessing the impact of additive social genetic effects on pig populations using records from automatic feeders. In this study, we used this approach to investigate the importance of additive social genetic effects in Pietrain and Iberian pigs. For the Pietrain population, additive social genetic effects accounted for a relatively small percentage of the phenotypic variance (approximately 1%). However, in the Iberian population, these effects played a more substantial role, explaining approximately 6.2% and 5.5% of the phenotypic variance for duration and feed intake per visit, respectively. Furthermore, we showed that the total heritable variation increased in Iberian pigs when additive social genetic effects were included, leading to a reduction in the residual variance.

The weak genetic correlation found between additive direct and additive social genetic effects at the visit level indicates independence between them. We suggested that high social genetic values are related to dominated or passive pigs, which was in line with the increase in the mean feed intake of the preceding pigs in the feeder. This study underscores the potential benefits of incorporating additive social genetic effects into breeding programmes. By effectively reducing residual variance, this inclusion improves the precision of breeding value estimates, facilitating enhanced response to selection. Additionally, this study contributes valuable insights into the genetic parameters of feeding behaviour traits in Iberian pigs within an intensive system, bridging the knowledge gap compared to studies focused on ‘white’ pig populations.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Data S1.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.