Variation in body weight and feed intake trajectories are promising resilience indicators in Texel lambs

Introduction: Resilient animals are capable of coping with environmental perturbations or quickly returning to unperturbed performance trajectory after facing challenges. More resilient animals tend to have better welfare, health, and productivity under variable conditions. However, trade-offs between production and resilience traits have been reported, indicating the need for further research to enable genetic selection for increased productive efficiency while maintaining or improving general resilience.

Methods: In this study, data from 76 Texel lambs monitored during a 53-day feed efficiency trial were used to generate 24 indicators of resilience based on variability in daily feed intake (FI), feeding behavior and average daily gain (ADG) and assess their phenotypic relationship with ADG and residual FI (RFI). Some traits evaluated included adgVar (residual variance of ADG), adgLnVar (log-variance of deviation between observed absolute and expected ADG), QRfi (quantile regression of FI), and QRdurfi (quantile regression of duration with effective consumption).

Results: Strong associations were found between indicators, such as adgVar and adgLnVar (r = 0.81). Productive traits showed two clear patterns, ADG was favorably correlated with QRdur (r = -0.53), QRdurfi (r = -0.65), QRfi (r = -0.65), suggesting that more resilient animals tend to have higher ADG. Conversely, RFI presented unfavorable correlations with resilience, ranging from r = -0.46 for QRfi to r = -0.24 for QRtimesfi indicating that more feed-efficient animals may be less resilient.

Discussion: These contrasting results highlight two key findings: (1) productivity and resilience can be favorably associated, as shown by ADG-resilience correlations, however, (2) specific feed efficiency indicators (e.g., RFI) may have antagonistic relationships with resilience. Given the relatively small sample size (n = 76) in this exploratory study, findings should be interpretated with caution but can provide some insights into the relation between resilience and production and potential trade-offs warranting further investigation.

1 Introduction

Animal resilience can be defined as the ability of animals to be minimally affected by environmental perturbations or, when affected, quickly recover and return to the state prior to the exposure to this perturbation (Colditz and Hine, 2016; Berghof et al., 2019). In the same line, robustness has been defined as the “animal’s ability to do the things it has to do to promote its future ability to reproduce” and can be characterized as a combination of resilience and adaptation (Friggens et al., 2017). So, while robustness is a consequence of longer-lasting environmental conditions, resilience tends to be expressed in response to environmental challenges that last for a shorter period such as few days (Colditz and Hine, 2016). The improvement in resilience will enhance the animals’ ability to adapt to changing environments and can still potentially reduce the effects of diseases or other stressors that impact animal welfare (Taghipoor et al., 2023; Wang et al., 2024).

A first step for increasing resilience through genetic selection is the definition of the indicator traits based on longitudinally recorded phenotypes (Berghof et al., 2019). The quantification of resilience at the individual level is needed for the inclusion of such indicator traits in breeding programs (Brito et al., 2020; Taghipoor et al., 2023). At the individual level, the dynamic properties of performance trajectories can provide insights into animal resilience, as variability in performance traits reflects the animal’s attempts to cope with the environmental challenges to which it is exposed (Colditz, 2022). Moreover, resilience indicators can be derived from the residuals estimated as the deviations between expected and observed performance (Berghof et al., 2019; Chen et al., 2023). High-frequency data (time-series data) is generally needed for accurately describing resilience. Such data enables the phenotyping of animals’ responses to perturbations and allows the quantification of the extent to which environmental disturbances elicit a response from the animal, as well as the speed at which recovery occurs (Friggens et al., 2017; 2022).

Precision livestock farming monitoring technologies have significantly increased the amount of data collected on commercial and experimental farms across production systems (Brito et al., 2020; Taghipoor et al., 2023). Longitudinal data on individual animals can provide detailed information on the animals’ health and welfare and their ability to cope with environmental disturbances (Elgersma et al., 2018; Taghipoor et al., 2023). For instance, feeders equipped with antennae and radio frequency identification tags can record individual feed intake and feeding behavior that can be used for deriving resilience traits (Putz et al., 2019; Graham et al., 2024). As resilience is a dynamic process influenced by multiple and complementary biological processes, complex modeling is needed such as dynamic mathematical models (Taghipoor et al., 2023) and multi-trait random regression models (Oliveira et al., 2019). Various researchers have proposed resilience indicators based on different data types and across livestock species, including data from automated milking systems in dairy cattle (e.g., Elgersma et al., 2018; Poppe et al., 2020; Chen et al., 2023), data from variation in body weight in poultry (Berghof et al., 2019), variability in feed intake (Putz et al., 2019) and vaginal temperature (Wen et al., 2024) in pigs, and variability in wool fiber diameter in sheep (Smith et al., 2024). These resilience indicators are designed to evaluate animals’ ability to cope with different environmental challenges, since they capture variation in performance and other physiological or behavioral indicators at the animal level. Some remarkable examples include those proposed by Berghof et al. (2019), such as the natural logarithm of variance, autocorrelation, and skewness of deviations, and the slope of reaction norm on specific environmental gradients applied to the body weight trajectory. Other examples are the ones proposed by Putz et al. (2019) such as quantile regression and root mean square error of feed intake in pigs. Both examples are centered in the hypothesis that feed intake, and consequently daily gain (ADG), are directly affected by environmental perturbations.

Although, to our knowledge, resilience indicators are not yet well-developed or included in meat sheep breeding programs, other traits are commonly used for selection decisions worldwide. For instance, body weight recorded at different ages are included in selection indexes from different breeds across countries (e.g., Texel in United Kingdom and Ireland). In a meta-analysis carried out by Mucha et al. (2022), resilience related traits such as those linked with disease resistance (e.g., mastitis, footroot, gastrointestinal parasites) were found to be uncorrelated or lowly correlated with efficiency traits. Although there is still limited information in this area, efficiency traits may not serve as auxiliary traits for improving animal resilience. Previous studies investigating the relationship between body weight traits with feed intake and ADG (Shrestha et al., 1985; Sinha and Singh, 1997; Snowder and Van Vleck, 2003) have reported moderate to high genetic correlations between ADG and feed intake with body weight traits (e.g., pre-weaning weight, post-weaning weight).

Intensive selection for greater productivity may have unfavorably affected animal robustness and resilience (Rauw et al., 1998; Douhard et al., 2021).Consequently, intensively selected animals tend to be more environmentally sensitive with a lower adaptive capacity when facing challenging environmental conditions (Misztal et al., 2025). Given that ADG has a moderate to high genetic correlation with growth traits, which have long been used in breeding programs, and residual feed intake (RFI) has been used for improving feed efficiency (Ferreira et al., 2024), there is a need for determining whether these productive efficiency traits are associated with general resilience.

Despite increasing interest in quantifying resilience in different species, studies focusing on sheep are still scarce, especially when combining feed intake, feeding behavior and growth-based indicators. The Texel breed was chosen as a model given its importance in the Uruguayan and worldwide sheep production sector. Texels are known for key traits such as high growth rate, leaner carcasses, high feed conversion efficiency and great muscling, which makes the breed an ideal population to explore potential trade-offs between production and resilience. In this context, the main objectives of this study were to derive indicators of overall resilience in Texel lambs based on variability in daily feed intake, feeding behavior, and ADG, and assess their relationship with ADG and RFI.

2 Materials and methods

2.1 Animals

All data were collected in the Intensive Sheep Phenotyping Platform located in the La Magnolia Experimental Unit (31°42’33.408”S, 55°49’38.172”W; Tacuarembó, Uruguay), which is part of the Instituto Nacional de Investigación Agropecuaria (INIA). In this experimental station, sheep are measured for feed intake, body weight, enteric methane emissions, and other relevant traits. Seventy-nine Texel lambs (36 females and 43 castrated males) with 6 to 7 months of age were measured for a period of 53 days. Animals were recorded during a feed efficiency test conducted from January 5 until February 26, 2023 (Summer period). The average daily temperature during this period was 24.41 °C degrees. The average maximum daily temperature was 31.21 °C degrees and the highest daily temperature observed during the recording period was 37.51 °C degrees. The average Temperature Humidity Index (THI) during the period was 72.74 ± 4.62 and the highest daily THI observed during the trial was 78.64. This period of 53 days also included 8 days of acclimatation, so resilience was derived based on information of 45 days. Although longer testing periods are desirable when deriving resilience indicators, similar or shorter periods have been used in other species (Wen et al., 2024; Graham et al., 2024). The animals used in this study came from the INIA Las Brujas information nucleus (Canelones, Uruguay) with an average test initial body weight of 38.07 ± 5.80 kg. These animals were weaned eight days prior to the beginning of the acclimatization period. The testing facility consisted of five 400 m² outdoor pens, which provided animals with access to shade (roof), and ad libitum feed and water. Animals were allocated to specific pens according to their initial body weight, to minimize the effects of dominance and social hierarchy.

Each pen was equipped with five automated feeders (Ponta^®, Betim, MG, Brazil) that have antennae readers for individually recognizing the animals at the feeder since each of them have a radio frequency identification (RFID) tag. Furthermore, the feeders were equipped with precision scales that generated information related to the amount of feed consumed on each visit. Besides the automated feeders, each pen contained two automated weighing platforms (Ponta^®, Betim, MG, Brazil) associated to the water drinkers for automatically recording body weight in each visit to the drinkers. All the data generated was sent to the Ponta software via a central computer, which allowed the monitoring of feed intake and body weight. Additionally, each animal was monitored daily, both visually by the staff and through the software that tracks the lambs’ access to the feeders and drinkers.

Animals were fed with lucerne haylage (dry matter 47.3%, crude protein 26.6%, neutral detergent fiber 31.0%, and acid detergent fiber 25.4%) fed ad libitum. More details on the feed efficiency test protocol can be found in Ferreira et al. (2021); Amarilho-Silveira et al. (2022), and (De Barbieri et al., 2024).

2.2 Statistical analyses

Regarding the body weight information, 27,115 raw records were available for the study. The quality control of these phenotypic records consisted of removing outliers and inconsistent or biologically implausible values. More specifically, we removed individual body weight records below 10 kg or above 130 kg. Moreover, prior to calculating the average daily weight per animal, outlier data points were removed, specifically those deviating by more than ±3 studentized residuals from the linear weight equation with the animal as a random effect. These steps reduced the body weight dataset to 26,715 observations. Concerning the feed intake phenotype, 392,770 raw records were initially available, the quality control for this phenotype was also based on biologically consistent valuables, data points for visit-based consumption below 0.00 kg or exceeding 2 kg were also removed. Additionally, consumptions higher than 1 kg in visits that lasted less than 3 minutes and consumptions greater than 1.5 kg within less than 4.5-minute visits were also removed. Visits that were above one hour of duration were removed, regardless of the consumption value. These steps resulted in a dataset with 392,666 observations used for the study. No missing data were present in the FI and ADG datasets. Therefore, there was no need for phenotypic data imputation or interpolation processes.

All statistical analyses were performed using the R software (R Core Team, 2023).

2.3 Feed intake phenotypes

The intakes per visit were summed up to generate the daily consumption of animals. Daily intake was transformed into analytical dry matter daily intake using a coefficient of 0.473, defined based on laboratory analysis of the forage for the whole test. A simple linear regression model was fitted for each animal, relating the analytical dry matter intake (ADMDI) to the feeding days, which were every day of the test. The equation used on this regression was as follows:

$\begin{array}{ll}ADMD{I}_{i\text{j}}={\beta }_{0i}+{\beta }_{1i}{X}_{i\text{j}}+{\text{ϵ}}_{\text{i}}& (1)\end{array}$

where $ADMDI$ = analytical dry matter intake of animal i on day j, ${\beta }_0$ = regression intercept for animal i, ${\beta }_1$ = slope of the regression line for animal i; and $X$ corresponds to the experimental day for animal i, and ϵ_i​ is the residual error term.

The linear regression analysis was performed using the lm() R function, and the residuals and fitted values were extracted with the augment() function of the broom package (Robinson et al., 2025). The adjustment of the linear regression was assessed individually for each animal. The quality of the adjustment was measured using the coefficient of determination (R²) and the root mean square error (RMSE), calculated as shown in Equation 2:

$\begin{array}{ll}RMSE= \sqrt{\frac{1}{n} {\displaystyle \sum _{i=1}^n}{(yi- \widehat{y}i)}^2 }& (2)\end{array}$

From the residuals, resilience indicators were derived following the approach proposed by Berghof et al. (2019), including skewness, the natural logarithm of the variance, and autocorrelation at lag one. Additional indicators were also calculated, such as the mean of negative residuals and the number of days with negative residuals. Skewness was extracted using the e1071 package (Meyer et al., 2024), and short-term temporal dependencies were assessed by computing autocorrelation at lag one using the acf() R function. For calculating the natural logarithm of the variance, the base R log() function, was used.

In addition to these traits, another set of resilience indicators were derived from the feed intake phenotype. Following the approach proposed by Putz et al. (2019), quantile regression (QR) was applied to feed intake (QRfi) and visit duration (QRdur) data. We also extended the analysis to include the duration of effective feeding (QRdurfi), the number of feeder visits per day (QRtimes, QRtimesfi), and feed intake per minute (QRfimin).

Quantile regression was applied to estimate ADMDI patterns across feeding days, allowing the identification of lower-than-expected intake events. To avoid penalizing smaller animals, feed intake data were corrected for body weight, since QR calculations take all observations into account (i.e., all feed consumption records for the animals studied). Expected feed intake values were estimated at the 10^th percentile to identify animals with consistently low intake. Observations falling below this threshold were flagged with a binary indicator representing off-feed days.

These off-feed days were aggregated within animals and classified then as a proportion of off-feed days, resulting in QR phenotypes, the higher the value (closer to 1) of this proportion, indicated more off-feed days the animal had. Each animal received one phenotype for the whole period.

Linear regressions were performed for both the duration at the feeder and the duration with effective feed consumption. These models were fitted over time to extract the root mean square error (RMSE), which was used as a resilience indicator. The resulting indicators were named RMSEdur and RMSEdurfi, respectively. This set of RMSE-based indicators also included RMSEfi, which had been previously described as a measure of the quality of the feed intake regression adjustment (see Equation 1).

The same methodology applied for the QRfi, QRdur, and QRdurfi was also used to the “times” variable, which is related to the number of daily visits to the feeder per animal and the also for the feed intake per minute in kilograms (fimin). To estimate these quantile regressions, the rqss() function from the quantreg package (Koenker, 2025) was used, as previously described. For QRfi, the regression was fitted using ADMDI as the response variable, with days and body weight as predictors.

The complete set of resilience indicators generated through the feed intake phenotype are shown in Table 1 along with a description of each indicator. These indicators are: fiVar, fiLnVar, fiSkew, fiNegN, fiNegSkew, fiACF, QRfi, QRdur, QRdurfi, QRfimin, QRtimes, QRtimesfi, RMSEfi, RMSEdur, and RMSEdurfi. After calculating the resilience indicators, descriptive statistics for each trait were computed to determine the mean, minimum, maximum, standard deviation, and coefficient of variation.

Table 1

Table 1. Framework and description of resilience indicators.

Residual feed intake (RFI) was calculated to later evaluate the phenotypic correlations between it and resilience traits, and the model for calculating RFI was:

$\begin{array}{ll}{Y}_{ijk}={\beta }_0+{\beta }_{1}B{W}_{ijk}^{0.75}+{\beta }_{3}AD{G}_{ijk}+{\beta }_4(Pe{n}_i x Se{x}_j)+e_{ijk}& (3)\end{array}$

where $Y_{ijk}$ = is the observed feed intake (sum of daily feed intake); ${\beta }_0$ is an all-animal constant that refers to the average daily feed intake; ${\beta }_{1}B{W}_{ijk}^{0.75}$ is the mid-test metabolic body weight (BW); ${\beta }_{3}AD{G}_{ijk}$ is the effect of ADG (described in the following section); ${\beta }_4(Pe{n}_{i}xSe{x}_j)$ is the interaction of pen (5 levels) with the sex of the animal; $e_{ijk}$ is the residual error indicating the RFI phenotypes (the difference between the observed and predicted feed intake).

2.4 Average daily gain phenotype

To evaluate bodyweight trajectories and deviations from expected patterns, for each animal, a linear regression of BW over time (days) was fitted to estimate the variable response: ADG during the 45-day data collection period. The regression was fitted using the lm() R function. ADG (slope) and the coefficient of determination (R²) were extracted using the broom package (Robinson et al., 2025) to assess the quality of the model fit. Additionally, the R² of ADG was later used as a resilience indicator. The regression model applied was as presented on Equation 4:

$\begin{array}{ll}Y= {\beta }_0+ {\beta }_{1}X+{\text{ϵ}}_{\text{i}}& (4)\end{array}$

where $Y$ = daily BW (in kg), ${\beta }_0$ = regression intercept, ${\beta }_1$ = coefficient of ADG (kg/day); and $X$ corresponds to the experimental day, and ϵ_i​ is the residual error term.

Residuals from this regression were extracted using the augment() function from the broom package (Robinson et al., 2025) and used to generate a set of resilience variables, as suggested by Berghof et al. (2019). These indicators were skewness (adgSkew) and the skewness of negative residuals (adgNegSkew), variance (adgVar), natural logarithm of variance (adgLnVar), mean of negative residuals (adgNeg), number of the days with negative residuals (adgrNegN) and the autocorrelation of the residuals (adgACF).

Skewness was calculated using the skewness() function from the e1071 R package, while the natural logarithm of the variance was obtained with the base R log() function. Short-term temporal dependencies of residuals were evaluated using the autocorrelation and at lag one, calculated with the acf() function in R. Additionally, the coefficient of determination (R²) were computed manually for each animal to assess the quality of the regression fit and used later as a resilience indicator.

This approach, based on analyzing residuals, assumes that more resilient animals exhibit smaller deviations from a healthy or steady state or performance potential, and recover faster (Sauvant and Martin, 2010; Doeschl-Wilson et al., 2021). In this context, this set of data is expected to capture when the animal is facing challenges, as they provide information on deviations from the expected growth trajectory. The ADG-derived indicators (Table 1) were: R^2ADG, adgVar, adgACF, adgLnVar, adgNeg, adgrNegN, adgSkew,adgNegSkew. After having the resilience indicators developed, the descriptive statistics of each trait were derived to establish the mean, minimum, maximum, standard deviation and coefficient of variation.

2.5 Phenotypic relationship among traits

After calculating the resilience indicators, Pearson correlation analyses were conducted to assess the relationship among all indicator traits using the rcorr function of the Hmisc R package (Harrell, 2025). Only statistically significant correlations were considered to report based on p-value threshold of 0.05. Detailed information, including correlation coefficients, confidence intervals, p-values, and significance levels are provided in the Supplementary Material.

3 Results

3.1 Descriptive statistics

Table 2 presents the descriptive statistics for all traits evaluated. The animals included in the study had an average BW of 38.07 ± 5.80 kg and an ADG of 0.19 ± 0.05 kg/d during the feeding test period. Individual ADG values ranged from -0.02 to 0.30 kg/d, indicating considerable variability among individuals and that some animals presented weight lost during the period. The high regression coefficient (0.92 ± 0.18) shows that ADG followed a linear pattern throughout the feeding test period.

Table 2

Table 2. Descriptive statistics of productive and resilience indicators.

Resilience-related traits based on ADG indicators presented phenotypic means ranging from -1.38 (adgLnVar) to 0.46 (adgACF) with coefficients of variation (CV) exceeding 30%, for all traits, except for adgrNegN which presented a lower CV of 10.94%. The adgrNegN trait had a mean of 21.37 ± 2.46, days (minimum 16; maximum 26) representing the days where the ADG was under the expected value. Similarly, the fiNegN had a mean value of 23.44 ± 2.42 days with a CV of 10.33% with a maximum value reaching 30 days. This implies that, within a 45-day trial, some animals spent up to 30 days of FI below the expected level. All QR indicators presented related means, ranging from 0.10 to 0.12, and similar (but high) CV, with values up to 196.03 (QRtimes), reinforcing the substantial individual differences.

3.2 Relationships among resilience indicators

Pearson correlations between the resilience indicators were analyzed to explore potential relations between them. The results are split into two groups: Figure 1 focuses on growth-related indicators and Figure 2 focuses on feeding-related indicators. The exact p-value and confidence intervals of the correlation’s coefficients are provided in the Supplementary Material section.

Figure 1

Figure 1. Phenotypic correlations between resilience indicators and productive traits: significant correlations for growth-related traits (p < 0.05 and p < 0.01). The list of abbreviations is presented in Table 1.

Figure 2

Figure 2. Phenotypic correlations between resilience indicators and productive traits: significant correlations for feeding-related traits (p < 0.05 and p < 0.01). The list of abbreviations is presented in Table 1.

The results revealed strong association between adgVar and adgLnVar (0.82); and between them and adgNeg (-0.90; -0.93). R²adg showed association with adgLnVar, adgVar, QRdur, QRdurfi, and QRfi, with coefficients of -0.60; -0.79; -0.56; -0.74; -0.70, respectively. The indicators fiSkew, fiNegN, fiNegSkew, QRtimes, and QRtimesfi were not associated with ADG-derived resilience indicators.

The largest significant correlations among FI-based indicators were between QRdur and QRdurfi (0.90), fiVar and RMSEfi (0.99), RMSEfi and fiLnVar (0.99), RMSEdur and RMSEdurfi (0.83), and fiLnVar and fiVar (0.95).

3.3 Pearson correlations of resilience indicators with ADG and RFI

Regarding correlation between resilience indicators and ADG, thirteen indicators showed to be significantly correlated with ADG, including adgLnVar, adgVar, adgNeg, adgACF, and R²adg, with correlation of -0.40, -0.49, 0.42, -0.31, and 0.74, respectively. The FI-based indicators significantly correlated with ADG were fiNegSkew, fiACF, QRdur, QRdurfi, QRfi, QRtimesfi, and RMSEdur, with correlations of -0.31, -0.31; -0.53, -0.65, -0.61, -0.25 and -0.24. These correlations are shown in Figure 1. Twelve resilience traits were statistically associated with RFI, including fiSkew (0.40), fiVar (0.48), fiLnVar (0.50), fiNeg (-0.36), fiNegN (0.39), QRdur (-0.23), QRfi (-0.46), QRfimin (-0.39), QRtimes (-0.24), QRtimesfi (-0.24), and RMSEfi (0.50).

4 Discussion

This study aimed to develop resilience indicators for potential use in Texel sheep breeding programs based on data collected during feeding trials. The proposed traits, based on measures of variability, autocorrelation, skewness, and quantile regression, are based on the premise that exposure to environmental perturbations typically lead to a reduction in feed intake, followed by decrease in ADG, and consequently, BW. These responses result in greater deviation between observed and predicted trajectory, as well as increased day-to-day variation in feeding behavior and growth. Such changes can be used to evaluate resilience, as animals experiencing stress are likely to temporarily reallocate resources from less-essential functions toward more crucial processes (Smith et al., 2024).

There is no universally accepted direct measure of resilience, and its quantification requires the use of mathematical models with biological reasoning, rather than solely relying on deviations between expected and observed performance (Berghof et al., 2019; Taghipoor et al., 2023). In this context, we evaluated resilience indicators previously described or proposed in the literature (Berghof et al., 2019; Putz et al., 2019). Variability in frequently recorded feed intake can provide valuable information regarding the health and welfare status of individual animals (Nguyen-Ba et al., 2020), as reduction in feed intake affects ADG and overall performance. The resilience indicators assessed in this study were developed under the hypothesis that animals exposed to environmental challenges consume less feed, spend less time at the feeder, and visit feeders less frequently compared to their typical feeding behavior. Many of the FI indicators evaluated in this study were previously proposed for pigs by Putz et al. (2019). Therefore, further investigation is needed to determine whether these indicators are also suitable for sheep.

Except for, R²adg, adgrNegN, fiLnVar, fiNegN, RMSEdur, and RMSEdurfi, all resilience indicators exhibited CV greater than 30%, ranging from 34.73 (RMSEfi) to 266.09 (fiACF). This reflects a high degree of phenotypic variability among animals, which may also indicate underlying genetic variability. Even for the QR-based indicators, which had similar mean values (Table 2), the high CVs indicate substantial data dispersion, highlighting individual differences in response patterns.

Residual variance reflects the extent to which an animal deviates from its expected trajectory, with higher variance indicating greater impacts of disturbances (Berghof et al., 2019). In this study, adgVar and fiVar had low average values (0.34 and 0.06, respectively), suggesting that, on average, animals did not experience large deviations during the testing period. This could be due to a certain degree of resilience within the population or the absence of significant environmental disturbances during the evaluation period. However, the high CVs for these indicators (137.76% for adgVar and 77.95% for fiVar) indicate that some animals had pronounced variability in performance, suggesting individual differences in resilience or in sensitivity to environmental or management-related stressors.

Skewness indicators were evaluated to assess whether deviations in performance were symmetrically distributed or biased toward negative or positive values, as this can provide insights into how animals respond to environmental challenges. According to Berghof et al. (2019), when an animal is not challenged or unaffected by a challenge, the skewness values are expected to be close to zero. The least favorable scenario occurs when skewness is negative, as this indicates that an animal experienced negative deviations in response to stressors. In the present study, mean values for adgSkew and fiSkew were –0.37 and 0.28, respectively. These results suggest that animals tended to exhibit downward deviations in ADG, potentially reflecting sensitivity to challenges that impair growth. In contrast, FI deviations were mostly positive, which may reflect compensatory intake behavior, as FI was more often above than below expected values. When skewness was calculated using the negative residuals (adgNegSkew; fiNegSkew), mean values were similar (-0.97 and -0.96, respectively). This finding suggests that when evaluating FI (fiSkew), general residuals are not very suitable and negative residuals may be more informative. The negative value of skewness observed for adgSkew (–0.37) aligns with the findings of Rodrigues et al. (2024) who also reported a negative average value for the BW skewness indicator in Guzerat cattle (Bos taurus indicus), although with a smaller magnitude (–0.08 ± 0.02). However, their results differed across two selection lines in Nellore cattle, where positive values were found for the same trait (0.06 ± 0.02 in one line, and 0.14 ± 0.02 in the other). This suggests genetic influence on resilience and growth-related traits.

The ACF values reflect the duration of a challenge’s impact on performance (Berghof et al., 2019). ACF values closer to zero suggest that the animal was either not exposed to environmental disturbances or recovered quickly from them. Conversely, ACF values closer to +1 are undesirable, as they indicate that the animal was affected by a challenge and experienced a slow recovery, evidenced by greater day-to-day variability in performance. In the present study, mean values for adgACF and fiACF were 0.45 and 0.06, respectively. At first glance, it may seem counterintuitive that FI exhibited lower autocorrelation than ADG, since FI is expected to show more immediate sensitivity to disturbances. However, the CV shows greater variability for fiACF (266.09%) compared to adgACF (36.3%), indicating considerable individual differences in FI autocorrelation. The QR indicators, developed specifically to assess FI-related traits, were based on the proportion of days classified as “off-feed” days, defined as days when residuals fell below a specified threshold (10% in this study). Animals with a greater proportion of such days had higher QR and were considered less resilient. The same principle applies to the RMSE indicators for FI: higher RMSE values indicate greater deviation from the expected trajectory and therefore lower resilience (Putz et al., 2019). Mean values of FI and feeding duration indicators, i.e. RMSEfi, RMSEdur, QRfi, and QRdur, were generally lower than those reported in previous studies (e.g., Putz et al., 2019; Chen et al., 2020), while part of this variation may be explained by inherent species differences in feeding behavior, anatomy, and physiology, as well as differences in experimental conditions. Specifically, we applied a 10% threshold for the QR calculations, whereas both Putz et al. (2019) and Chen et al. (2020) used a more restrictive 5% threshold. The stricter threshold likely identified fewer off-feed events, resulting in lower QR values in their studies. We initially implemented the 5% threshold as suggested by ​Putz et al. (2019).​ However, this value was found to be overly restrictive in our case, as the animals were not subjected to an artificial disease challenge, and the environment closely resembled the conditions encountered by animals raised in extensive systems. Consequently, we adopted a 10% threshold. Furthermore, the study by Putz et al. (2019) was conducted in pigs under disease challenge scenarios, primarily to assess disease resilience, whereas the present study was carried out without artificial challenges which may have contributed to differences in the observed resilience indicators, as compared with previous literature.

To explore the flexibility and broader applicability of the resilience framework proposed by Putz et al. (2019), we extended their methodology, originally applied to FI and feeding duration (RMSEfi, RMSEdur, QRfi, QRdur) to additional phenotypes. This resulted in five new resilience indicators, which were QRdurfi, RMSEdurfi, QRfimin, QRtimes, and QRtimesfi. These variables were selected based on the detailed and high-frequency data available from the intensive phenotyping platform, which allows for the disaggregation of feeding behavior into time-based and frequency-based components. The mean value of these last indicators was respectively 0.11 (QRdurfi), 27.40 (RMSEdurfi), 0.10 (QRfimin), 0.10 (QRtimes), and 0.10 (QRtimesfi), which were closer to what was found for the indicators derived from Putz et al. (2019). These results suggests that the resilience quantification approach through these phenotypes can be applied to an extended range of indicators offering a more comprehensive view of an animal’s response and behavior to perturbations.

4.1 Pearson correlations

As previously discussed, many resilience indicators were not significantly correlated among themselves and the significant associations that did emerge were generally weak to moderate (Figures 1, 2). This suggests that each indicator may be capturing different aspects of the animals’ response to challenges. Among the stronger correlations adgLnVar and adgVar showed a high positive correlation (0.82), indicating that these two measures reflect similar aspects of variability in ADG. Both indicators (adgLnVar and adgVar) were also favorably correlated with R²adg (-0.60; -0.79) supporting the interpretation that greater variability in residuals corresponds to less predictable growth patterns. The correlation between adgLnVar and adgVar with adgNeg (-0.93 and -0.90) may be counterintuitive but could indicate that animals with greater overall variability tend to be more consistent when evaluated under negative residuals. In addition, adgSkew was favorably correlated with adgNegN and adgNegSkew (0.53; 0.49), indicating that changes in the overall skewness of the residuals are reflected in the pattern of the negative residuals. This suggests that skewness indicators, especially when applied to the negative residuals may be useful for capturing asymmetries in individual performance under perturbations.

The adgACF indicator was significantly correlated with three other indicators (R²adg QRdurfi and QRfi) with coefficients of -0,37; 0.27 and 0.22, this agrees with the observations of Berghof et al. (2019), who proposed that higher variance is associated with lower resilience. Animals with greater variance tend to show increased day-to-day fluctuations in performance, which is reflected in higher autocorrelation. Further evaluation using a larger number of animals is needed to better understand the applicability of ACF indicators to other traits. The fiACF indicator showed significant correlations with other three traits (QRfi, QRdurfi, QRdur) with coefficients of 0,36; 0,42; 0,34, in agreement with what was previously sustained, where higher autocorrelation is accompanied by higher variation, in this case, higher QR’s. The fiLnVar was strongly and positively correlated with RMSEfi (r = 0.99), supporting the findings of Putz et al. (2019), who reported that less resilient animals exhibited greater deviation from expected feed intake patterns, as reflected by both higher variance and larger RMSE values.

Some correlations involving fiLnVar did not align with expectations based on previous literature. For instance, fiLnVar had a phenotypic correlation of 0.49 with fiSkew. Less resilient animals should present greater variance, and it should be associated with more negative skewness, reflecting a predominance of negative deviations. However, the positive skewness observed in this study suggests that some animals might have shown period of compensatory FI, possibly reflecting some period of short-term recoveries. These results may indicate that fiLnVar is primarily capturing the magnitude of positive deviations, which is reflected in the positive asymmetry (positive skewness) of the residual distribution. However, when evaluating the correlation between fiLnVar and fiNegSkew (which accounts only for negative residual variance) it was not observed significantly correlation coefficient. The small sample size (number of animals) should be improved in future studies to assess whether (1) accounting for the negative residuals is a suitable approach, and (2) whether a consistent relationship exists between fiNegSkew, fiVar, and fiLnVar.

The R²adg trait was significantly correlated with some QR traits, showing coefficients of –0.56, –0.74, and –0.70 with QRdur, QRdurfi, and QRfi, respectively. This indicates that more resilient animals (higher R²adg) tend to have lower QR values as a result of higher consistency at the feeder or maintaining overall feed intake when facing challenging conditions. This observed pattern reinforces that resilient animals maintain performance with less aversive effects in feeding behavior. The RMSEfi indicator showed a correlation of -0.88 with fiNeg. A negative correlation between RMSE and the variance of negative residuals suggests that individuals with higher overall error (higher RMSE) tend to show more consistent deviations below their expected trajectory, whereas individuals with lower RMSE display more variable but generally smaller negative deviations. This pattern might indicate that less resilient animals experience persistently lower performance, while more resilient ones maintain overall stability but can occasionally exhibit variable short-term declines.

In relation to the connection between production and resilience, Ramón et al. (2021) emphasized that, in the context of climate change and the goal of breeding for improved climate stress resilience, the use of novel traits more directly linked to animal resilience could be advantageous. However, they also cautioned about the potential antagonism between resilience and production traits, which must be considered when developing or refining selection indexes. Previous studies have also reported antagonistic relationships between production and robustness traits (Rauw et al., 1998; Maskal et al., 2024). The Pearson correlation between resilience indicators and between productive traits (ADG and RFI) in our study showed that resilience is mostly uncorrelated or weakly correlated with ADG or RFI at the phenotypic level. These findings indicate that ADG and RFI themselves, as evaluated in the current study, may provide limited information about animal resilience. Consequently, current selection for higher productivity and efficiency is not expected to result in high indirect responses on animal resilience. However, this needs to be confirmed with genetic correlation analyses and using larger datasets.

Concerning the discussion of potential antagonisms, the correlations between ADG and the QR indicators (QRdur, QRdurfi and QRfi) did not support the hypothesis of a negative relationship between productivity and resilience. According to these results, more productive animals tended to show lower QR values, suggesting that high growth rates can coexist with stable feeding behavior even when the animal is facing some perturbation. This can lead to a discussion that the more productive animals are not necessarily less resilient at the phenotypic level. When observing the correlation between RFI and resilience indicators, no strong correlation was observed, indicating a possible lack of association between these two groups of traits. However, even though weak to moderate correlations were observed between RFI and QR indicators (QRdur, QRfi, QRfimin, QRtimes, QRtimesfi), the opposite was observed compared to ADG. In this case, the negative correlations (-0.23; -0.46; -0.39; -0.24; -0.24) were unfavorable, meaning that higher RFI (i.e., less efficient animals) had lower QR’s (i.e., more resilient animals). Conversely, RFI was positively correlated with RMSEfi and fiLnVar (0.50 and 0.50, respectively), suggesting that less feed efficient animals also exhibited greater variation in FI trajectory (i.e., lower resilience). This apparent inconsistency likely reflects that QR indicators and RMSEfi capture different aspects of feeding behavior stability. While QR indicators represent day-to-day variability and regularity in intake, RMSE reflects the overall predictability of the individual feeding trajectory. Therefore, efficient animals (low RFI) may show more stable and regular feeding patterns, yet deviations from the modeled trajectory may still occur under specific conditions, leading to higher RMSE values. Moreover, Douhard et al. (2022) compared different lines (resistant vs. susceptible to parasites, and more and less feed efficient) of Romane sheep for potential differences in energy allocation. No trade-offs were observed reflecting that there was no conflict between efficiency and resistance, and in some cases, there was a positive relation between these traits, where more efficient animals were also more resistant. These results highlight the complexity of the relationship between efficiency, productivity, and resilience, and reinforce the necessity for further analyses assessing their genetic relationships using larger datasets. Although ADG is already incorporated into ADG-derived indicators, the information that these indicators and resilience are positively associated can also be resulted since animals in better condition tend to be more stable in their performance under some kind of challenges (Cornelius et al., 2014). On the other hand, the indication that feed efficiency shows an opposite relationship to resilience highlights a discussion about the current selection goals focused on productive efficiency. Since feed efficiency is a key trait to farmers (Pravia et al., 2022), it is important to consider whether this focus might be unintentionally leading to the selection of animals that are less resilient to short-term environmental disturbances.

This finding of association between productive traits (ADG this case) and some resilience indicators were explored in previous dairy cattle studies. Keßler et al. (2024) reported phenotypic correlations of milk yield (MY) with resilience indicators in dairy cattle such as variance and natural logarithm of variance of 0.12 and 0.14, respectively. These results indicate a low association between MY and resilience, indicating that cows with higher MY tend to be less resilient. Additionally, Poppe et al. (2020) reported low phenotypic correlations between resilience indicators (LnVar, autocorrelation and skewness) based on negative deviations from different curves of lactation in dairy cattle, and average daily MY (AMY). The correlations between AMY and autocorrelation ranged from 0.07 and 0.11, from -0.05 to 0.01 between AMY and skewness, and from 0.10 to 0.27 between AMY and natural logarithm of variance. Further evidence of trade-offs between production and adaptative or coping traits has been presented. Maskal et al. (2024) performed a meta-analysis for heritability and genetic correlation for resilience and productive indicators in Holstein cattle, and reported genetic correlations ranging from -0.360 ± 0.25 (protein content and milk acetone concentration) to 0.535 ± 0.72 (measures related to fat-to-protein ratio vs. milk acetone concentration). The favorable and unfavorable genetic correlations observed reinforce the complexity of these traits and the need for taking this information into account when refining selection indices.

Haga clic o pulse aquí para escribir texto.

Although this study presents an important first step, it would be beneficial to apply the same methodology to larger datasets and estimate variance components and genetic correlations between trait pairs to better understand the relationship between resilience and production efficiency traits in Texel sheep. Few previous studies on resilience in sheep have provided encouraging evidence regarding the feasibility of evaluating and then selecting for this trait in this species. For example, Garcia-Baccino et al. (2021) explored a data driven approach to estimate the probability that a given day an environmental challenge occurred and explored the genetic determinism of resilience to these events using daily feed intake records. Their results revealed genotype-by-environment interaction which indicates that animals performing well in stable environments may not maintain the same level in harsher environments. Moreover, previous studies in other species, such as dairy cattle, have assessed genetic parameters of resilience and have shown that these indicators are heritable (Maskal et al., 2024; Brito et al., 2025) which highlights the opportunities for future studies and the possibility of including this trait in future breeding programs. Together with our findings, which revealed high variability among animals, these previous studies support the evaluation of animal resilience and reinforce the possibility of including it in future selection indices.

The lack of strong correlations among resilience indicators in this study suggests that general resilience is a multifaceted trait that must be captured through multiple variables reflecting different biological mechanisms. Additionally, the animals evaluated may not have experienced major environmental disturbances during the testing period. While some recent studies on animal resilience have analyzed larger datasets (Garcia-Baccino et al., 2021; Graham et al., 2024; Putz et al., 2019) others have applied alternative methodologies for assessing resilience and welfare proxies in a similar number of animals (Bergeron et al., 2024; Queiroz de Carvalho et al., 2024; Sartori et al., 2024) .

This study was conducted under conditions closely resembling extensive systems, without any artificial perturbations imposed on the animals, which may influence the observed results. Future research could evaluate the animals under more challenging environmental conditions, such as periods of heat waves, heavy precipitation, or human interventions.

5 Conclusions

The increasing availability of high-frequency data provides new opportunities for exploring animal resilience. Variability in ADG and FI can provide information on animal health and welfare status and can be used for deriving resilience indicators to capture the animals’ response to environmental perturbations. In the present study, the evaluated resilience indicators showed complex interactions both among themselves and with productive traits. The correlations between production and resilience suggest that the current selection for greater body growth rate might not have substantially affected animal resilience. However, correlations between feed efficiency and resilience ranged from favorable to unfavorable, suggesting that selection for improved feed efficiency could either enhance or compromise certain resilience indicators. Although the relatively low sample size in this study limits the strength of these conclusions, this methodology will be applied to larger datasets to further validate the findings. The use of high-frequency phenotypes such as feed intake and average daily gain could help to identify animals that maintain more stable performance under variable conditions. Considering that resilience and production can sometimes follow different directions, integrating resilience indicators in multi-trait selection indices represents an important opportunity to select animals that are more capable to sustain productivity while coping with environmental challenges. Future research should focus on investigating the genetic relationships between these resilience indicators and production traits while exploring the feasibility of including resilience traits in sheep breeding programs.

Data availability statement

The datasets presented in this article are not readily available because The datasets presented in this article are not publicly available due to legal restrictions. The data were obtained by INIA and can be made available for research purposes from the authors with permission from INIA. Requests to access the datasets should be directed to Ignacio De Barbieri, aWRlYmFyYmllcmlAaW5pYS5vcmcudXk=.

Ethics statement

The animal study was approved by INIA Animal Ethics Committee, with protocol number INIA 2018.2. The study was conducted in accordance with the local legislation and institutional requirements.

Author contributions

FB: Formal analysis, Investigation, Methodology, Writing – original draft, Writing – review & editing. GC: Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Methodology, Writing – review & editing. LB: Investigation, Methodology, Writing – review & editing. ID: Conceptualization, Funding acquisition, Investigation, Methodology, Project administration, Writing – review & editing.

Funding

The author(s) declared that financial support was received for this work and/or its publication. The authors declare that financial support was received from the Agencia Nacional de Investigación e Innovación (Uruguay), with code POS_NAC_2023_2_177523, and from INIA for funding Fernanda Barchet’s MSc scholarship. This study was supported by GENERA project (ARN_18) funded by INIA.

Acknowledgments

The authors would like to acknowledge INIA staff, particularly those located at the La Magnolia Experimental Unit (Tacuarembó, Uruguay) for their work and commitment on carrying out the efficiency test.

Conflict of interest

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

The author(s) declared that they were an editorial board member of Frontiers, at the time of submission. This had no impact on the peer review process and the final decision.

Generative AI statement

The author(s) declare that Generative AI was used in the creation of this manuscript. ChatGPT (version GPT-4, model OpenAI GPT-4-turbo, developed and maintained by OpenAI) was used to help with the translation of the original draft and to improve the readability of the manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Supplementary material

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fanim.2025.1688769/full#supplementary-material

References

Amarilho-Silveira F., de Barbieri I., Cobuci J. A., Balconi C. M., de Ferreira G. F., and Ciappesoni G. (2022). Residual feed intake for Australian Merino sheep estimated in less than 42 days of trial. Livestock Sci. 258, 104889. doi: 10.1016/j.livsci.2022.104889

Crossref Full Text | Google Scholar

Bergeron J. H., Huber S., Shane T., Brito L. F., Washington-Allen R., and Hess A. (2024). Using GPS collars and vaginal temperature sensors for characterizing rangeland usage and climatic stress in sheep. J. Anim. Sci. Abstracts. 102, 402. doi: 10.1093/jas/skae234.458

Crossref Full Text | Google Scholar

Berghof T., Poppe M., and Mulder H. (2019). Opportunities to improve resilience in animal breeding programs. Front. Genet. 10. doi: 10.3389/fgene.2018.00692

PubMed Abstract | Crossref Full Text | Google Scholar

Brito L. F., Klaas I. C., Schodl K., Cabrera V. E., Stygar A., Iwersen M., et al. (2025). Invited Review: Using data from sensors and other precision farming technologies to enhance the sustainability of dairy cattle breeding programs. J. Dairy Sci. 108, 10447–10474. doi: 10.3168/jds.2025-26554

PubMed Abstract | Crossref Full Text | Google Scholar

Brito L. F., Oliveira H. R., McConn B. R., Schinckel A. P., Arrazola A., Marchant-Forde J. N., et al. (2020). Large-scale phenotyping of livestock welfare in commercial production systems: A new frontier in animal breeding. Front. Genet. 11. doi: 10.3389/fgene.2020.00793

PubMed Abstract | Crossref Full Text | Google Scholar

Chen S., Boerman J., Gloria L., Pedrosa V., Doucette J., and Brito L. (2023). Genomic-based genetic parameters for resilience across lactations in North American Holstein cattle based on variability in daily milk yield records. J. Dairy Sci. 106, 4133–4146. doi: 10.3168/jds.2022-22754

PubMed Abstract | Crossref Full Text | Google Scholar

Chen Y., Tibbs Cortes L. E., Ashley C., Putz A. M., Lim K. S., Dyck M. K., et al. (2020). The genetic basis of natural antibody titers of young healthy pigs and relationships with disease resilience. BMC Genomics 21, 1–14. doi: 10.1186/s12864-020-06994-0

PubMed Abstract | Crossref Full Text | Google Scholar

Colditz I. G. (2022). Competence to thrive: Resilience as an indicator of positive health and positive welfare in animals. Anim. Production Sci. 62, 1439–1458. doi: 10.1071/AN22061

Crossref Full Text | Google Scholar

Colditz I. G. and Hine B. C. (2016). Resilience in farm animals: Biology, management, breeding and implications for animal welfare. Anim. Production Sci. 56, 1961–1983. doi: 10.1071/AN15297

Crossref Full Text | Google Scholar

Cornelius M. P., Jacobson C., and Besier R. B. (2014). Body condition score as a selection tool for targeted selective treatment-based nematode control strategies in Merino ewes. Veterinary Parasitol. 206, 173–181. doi: 10.1016/j.vetpar.2014.10.031

PubMed Abstract | Crossref Full Text | Google Scholar

De Barbieri I., Navajas E. A., Ramos Z., Ferreira G., Velazco J., and Ciappesoni G. (2024). Feed conversion efficiency does not negatively affect young sheep and ewe performance. Front. Anim. Sci. 5. doi: 10.3389/fanim.2024.1480928

Crossref Full Text | Google Scholar

Doeschl-Wilson A., Knap P. W., Opriessnig T., and More S. J. (2021). Review: Livestock disease resilience: From individual to herd level. Animal 15, 100286. doi: 10.1016/j.animal.2021.100286

PubMed Abstract | Crossref Full Text | Google Scholar

Douhard F., Douhard M., Gilbert H., Monget P., Gaillard J. M., and Lemaître J. F. (2021). How much energetic trade-offs limit selection? Insights from livestock and related laboratory model species. Evolutionary Appl. 14, 2726–2749. doi: 10.1111/eva.13320

PubMed Abstract | Crossref Full Text | Google Scholar

Douhard F., Rupp R., and Gilbert H. (2022). “Feed efficiency and resource allocation trade-offs: theory, evidence and prospects,” in Proceedings of 12th World Congress on Genetics Applied to Livestock Production (WCGALP) Technical and species orientated innovations in animal breeding, and contribution of genetics to solving societal challenges (Wageningen Academic Publishers, Wageningen, Netherlands), 264–267.

Google Scholar

Elgersma G. G., de Jong G., van der Linde R., and Mulder H. A. (2018). Fluctuations in milk yield are heritable and can be used as a resilience indicator to breed healthy cows. J. Dairy Sci. 101, 1240–1250. doi: 10.3168/jds.2017-13270

PubMed Abstract | Crossref Full Text | Google Scholar

Ferreira G. F., Ciappesoni G., Castells D., Amarilho-Silveira F., Navajas E. A., Giorello D., et al. (2021). Feed conversion efficiency in sheep genetically selected for resistance to gastrointestinal nematodes. Anim. Production Sci. 61, 754–760. doi: 10.1071/AN20121

Crossref Full Text | Google Scholar

Ferreira J., Chay-Canul A. J., De Barbieri I., and da Costa R. L. D. (2024). Compilations and updates on residual feed intake in sheep. Trop. Anim. Health Production 56, 75–760. doi: 10.1007/s11250-024-04018-7

PubMed Abstract | Crossref Full Text | Google Scholar

Friggens N. C., Adriaens I., Boré R., Cozzi G., Jurquet J., Kamphuis C., et al. (2022). Resilience: Reference measures based on longer-term consequences are needed to unlock the potential of precision livestock farming technologies for quantifying this trait. Peer Community J. 2, 136. doi: 10.24072/pcjournal.136

Crossref Full Text | Google Scholar

Friggens N., Blanc F., Berry D., and Puillet L. (2017). Review: Deciphering animal robustness. A synthesis to facilitate its use in livestock breeding and management. Animal 11, 2237–2251. doi: 10.1017/S175173111700088X

PubMed Abstract | Crossref Full Text | Google Scholar

Garcia-Baccino C. A., Marie-Etancelin C., Tortereau F., Marcon D., Weisbecker J. L., and Legarra A. (2021). Detection of unrecorded environmental challenges in high-frequency recorded traits, and genetic determinism of resilience to challenge, with an application on feed intake in lambs. Genet. Selection Evol. 53, 1–14. doi: 10.1186/s12711-020-00595-x

PubMed Abstract | Crossref Full Text | Google Scholar

Graham J. R., Taghipoor M., Gloria L. S., Boerman J. P., Doucette J., Rocha A. O., et al. (2024). Trait development and genetic parameters of resilience indicators based on variability in milk consumption recorded by automated milk feeders in North American Holstein calves. J. Dairy Sci. 107, 11180-11194. doi: 10.3168/jds.2024-25192

PubMed Abstract | Crossref Full Text | Google Scholar

Harrell F. (2025). Hmisc: Harrell miscellaneous (Version 5.2-3) (Comprehensive R Archive Network (CRAN). doi: 10.32614/CRAN.package.Hmisc

Crossref Full Text | Google Scholar

Keßler F., Wellmann R., Chagunda M. G. G., and Bennewitz J. (2024). Resilience indicator traits in 3 dairy cattle breeds in Baden-Württemberg. J. Dairy Sci. 107, 3780–3793. doi: 10.3168/jds.2023-24305

PubMed Abstract | Crossref Full Text | Google Scholar

Koenker R. (2025). quantreg: Quantile regression (Version 6.1) (Comprehensive R Archive Network (CRAN). doi: 10.32614/CRAN.package.quantreg

Crossref Full Text | Google Scholar

Maskal J. M., Pedrosa V. B., Rojas de Oliveira H., and Brito L. F. (2024). A comprehensive meta-analysis of genetic parameters for resilience and productivity indicator traits in Holstein cattle. J. Dairy Sci. 107, 3062–3079. doi: 10.3168/jds.2023-23668

PubMed Abstract | Crossref Full Text | Google Scholar

Meyer D., Dimitriadou E., Hornik K., Weingessel A., and Leisch F. (2024). e1071: Misc functions of the Department of Statistics, Probability Theory Group (formerly: E1071) (Version 1.7-16) (Comprehensive R Archive Network (CRAN). doi: 10.32614/CRAN.package.e1071

Crossref Full Text | Google Scholar

Misztal I., Brito L. F., and Lourenco D. (2025). Breeding for improved heat tolerance in dairy cattle: methods, challenges, and progress. JDS Communications 6(3), 464–468. doi: 10.3168/jdsc.2024-0651

PubMed Abstract | Crossref Full Text | Google Scholar

Mucha S., Tortereau F., Doeschl-Wilson A., Rupp R., and Conington J. (2022). Animal Board Invited Review: Meta-analysis of genetic parameters for resilience and efficiency traits in goats and sheep. Animal 16, 100456. doi: 10.1016/j.animal.2022.100456

PubMed Abstract | Crossref Full Text | Google Scholar

Nguyen-Ba H., Van Milgen J., and Taghipoor M. (2020). A procedure to quantify the feed intake response of growing pigs to perturbations. Animal 14, 253–260. doi: 10.1017/S1751731119001976

PubMed Abstract | Crossref Full Text | Google Scholar

Oliveira H. R., Brito L. F., Lourenco D. A. L., Silva F. F., Jamrozik J., Schaeffer L. R., et al. (2019). Invited review: Advances and applications of random regression models: From quantitative genetics to genomics. J. Dairy Sci. 102, 7664–7683. doi: 10.3168/jds.2019-16265

PubMed Abstract | Crossref Full Text | Google Scholar

Poppe M., Veerkamp R. F., van Pelt M. L., and Mulder H. A. (2020). Exploration of variance, autocorrelation, and skewness of deviations from lactation curves as resilience indicators for breeding. J. Dairy Sci. 103, 1667–1684. doi: 10.3168/jds.2019-17290

PubMed Abstract | Crossref Full Text | Google Scholar

Pravia M. I., Navajas E. A., Aguilar I., and Ravagnolo O. (2022). Evaluation of feed efficiency traits in different Hereford populations and their effect on variance component estimation. Anim. Production Sci. 62, 1652–1660. doi: 10.1071/AN21420

Crossref Full Text | Google Scholar

Putz A. M., Harding J. C. S., Dyck M. K., Fortin F., Plastow G. S., and Dekkers J. C. M. (2019). Novel resilience phenotypes using feed intake data from a natural disease challenge model in wean-to-finish pigs. Front. Genet. 10. doi: 10.3389/fgene.2018.00660

PubMed Abstract | Crossref Full Text | Google Scholar

Queiroz de Carvalho D. T., Marques Ferreira B. J., Matos J. C., Nascimento Ramos E. J., Gois G. C., Leandro de Carvalho F. A., et al. (2024). Interaction between residual feed intake and thermal environment on performance, nitrogen balance, ingestive behavior and carcass yield of dorper lambs. J. Thermal Biol. 119, 103802. doi: 10.1016/j.jtherbio.2024.103802

PubMed Abstract | Crossref Full Text | Google Scholar

R Core Team (2023). R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria. Available online at: https://www.R-project.org/ (April 16, 2025).

Google Scholar

Ramón M., Carabaño M. J., Díaz C., Kapsona V. V., Banos G., and Sánchez-Molano E. (2021). Breeding strategies for weather resilience in small ruminants in Atlantic and Mediterranean climates. Front. Genet. 12. doi: 10.3389/fgene.2021.692121

PubMed Abstract | Crossref Full Text | Google Scholar

Rauw W. M., Kanis E., Noordhuizen-Stassen E. N., and Grommers F. J. (1998). Undesirable side effects of selection for high production efficiency in farm animals: A review. Livestock Production Sci. 56, 15–33. doi: 10.1016/S0301-6226(98)00147-X

Crossref Full Text | Google Scholar

Robinson D., Hayes A., and Couch S. (2025). broom: Convert statistical objects into tidy tibbles (Version 1.0.8) (Comprehensive R Archive Network (CRAN). doi: 10.32614/CRAN.package.broom

Crossref Full Text | Google Scholar

Rodrigues G. R. D., Rezende V. T., Mercadante M. E. Z., Bonilha S. F. M., Canesin R. C., Raineri C., et al. (2024). Animal growth models as a tool to estimate resilience indicators in Bos indicus and Bos taurus heifers: Selection effects and genetic parameters. Livestock Sci. 282, 105435. doi: 10.1016/j.livsci.2024.105435

Crossref Full Text | Google Scholar

Sartori L. V. C., Gurgeira D. N., Crisóstomo C., de Paz C. C. P., Ferreira J., and da Costa R. L. D. (2024). Performance, feed efficiency, feeding behavior, and cortisol concentration of lambs ranked for residual feed intake and residual intake and gain. Trop. Anim. Health Production 56. doi: 10.1007/s11250-024-03894-3

PubMed Abstract | Crossref Full Text | Google Scholar

Sauvant D. and Martin O. (2010). Robustesse, rusticité, flexibilité, plasticité: Les nouveaux critères de qualité des animaux et des systèmes d’élevage. INRAE Productions Animales 23, 5–10. doi: 10.20870/productions-animales.2010.23.1.3280

Crossref Full Text | Google Scholar

Shrestha J. N. B., Vesely J. A., and Chesnais J. P. (1985). Genetic and phenotypic parameters for daily gain and body weights in Suffolk lambs. Canadian Journal of Animal Science 65, 575-582. doi: 10.4141/cjas85-068

Crossref Full Text | Google Scholar

Sinha N. K. and Singh S. K. (1997). Genetic and phenotypic parameters of body weights, average daily gains and first shearing wool yield in Muzaffarnagri sheep. Small Ruminant Res. 26, 203–211. doi: 10.1016/S0921-4488(97)00025-0

Crossref Full Text | Google Scholar

Smith E. G., Walkom S. F., and Clark S. A. (2024). Exploring genetic variation in potential indicators of resilience in sheep using fibre diameter measured along the wool staple. Animal 18, 101065. doi: 10.1016/j.animal.2023.101065

PubMed Abstract | Crossref Full Text | Google Scholar

Snowder G. D. and Van Vleck L. D. (2003). Estimates of genetic parameters and selection strategies to improve the economic efficiency of postweaning growth in lambs. J. Anim. Sci. 81, 2704–2716. doi: 10.2527/2003.81112704x

PubMed Abstract | Crossref Full Text | Google Scholar

Taghipoor M., Pastell M., Martin O., Nguyen Ba H., van Milgen J., Doeschl-Wilson A., et al. (2023). Animal board invited review: Quantification of resilience in farm animals. Animal 17, 100925. doi: 10.1016/j.animal.2023.100925

PubMed Abstract | Crossref Full Text | Google Scholar

Wang A., Su G., Brito L. F., Zhang H., Shi R., Liu D., et al. (2024). Investigating the relationship between fluctuations in daily milk yield as resilience indicators and health traits in Holstein cattle. Journal of Dairy Science 107, 1535–1548. doi: 10.3168/jds.2023-23495

PubMed Abstract | Crossref Full Text | Google Scholar

Wen H., Johnson J. S., Gloria L. S., Araujo A. C., Maskal J. M., Hartman S. O., et al. (2024). Genetic parameters for novel climatic resilience indicators derived from automatically-recorded vaginal temperature in lactating sows under heat stress conditions. Genet. Selection Evol. 56, 1–19. doi: 10.1186/s12711-024-00908-4

PubMed Abstract | Crossref Full Text | Google Scholar