Capturing resilience from phenotypic deviations: a case study using feed consumption and whole genome data in pigs

Abstract

Background: In recent years, interest has grown in quantifying resilience in livestock by examining deviations in target phenotypes. This method is based on the idea that variability in these phenotypes reflects an animal's ability to adapt to external factors. By utilizing routinely collected time-series feed intake data in pigs, researchers can obtain a broad measure of resilience. This measure extends beyond specific conditions, capturing the impact of various unknown external factors that influence phenotype variations. Importantly, this method does not require additional phenotyping investments. Despite growing interest, the relationship between resilience indicators-calculated as deviations from longitudinally recorded target traits-and the mean of those traits remains largely unexplored. This gap raises the risk of inadvertently selecting for the mean rather than accurately capturing true resilience. Additionally, distinguishing between random phenotype fluctuations (white noise) and structural variations linked to resilience poses a challenge. With the aim of developing general resilience indicators applicable to commercial swine populations, we devised four resilience indicators utilizing daily feed consumption as the target trait. These include a canonical resilience indicator (BALnVar) and three novel ones (BAMaxArea, SPLnVar, and SPMaxArea), designed to minimize noise and ensure independence from daily feed consumption. We subsequently integrated these indicators with Whole Genome Sequencing using SLEMM algorithm, data from 1,250 animals to assess their efficacy in capturing resilience and their independence from the mean of daily feed consumption.

Results: Our findings revealed that conventional resilience indicators failed to differentiate from the mean of daily feed consumption, underscoring potential limitations in accurately capturing true resilience. Notably, significant associations involving conventional resilience indicators were identified on chromosome 1, which is commonly linked to body weight.

Conclusion: We observed that deviations in feed consumption can effectively serve as indicators for selecting resilience in commercial pig farming, as confirmed by the identification of genes such as PKN1 and GYPC. However, the identification of other genes, such as RNF152, related to growth, suggests that common resilience quantification methods may be more closely related to the mean of daily feed consumption rather than capturing true resilience.

Keywords: Environmental variance; Feed consumption; LnVar; Resilience; Swine; Whole genome sequence.

Fig. 1

(A) Deviation of observed feed consumption deviance (FCD): This panel shows the deviation of observed FCD (white points and light blue lines) compared to the expected FCD calculated using linear regression (red line) and spline regression (dark blue line). (B) Residuals of BA Model: This panel illustrates the residuals from the Base Adjustment (BA) model, representing the differences between the observed FCD and the estimated FCD from linear regression (red line) The red shaded area highlights the periods under the curve for each negative period. (C) Residual Differences: This panel shows the residual differences between spline regression (dark blue line) and linear regression (red line). The red shaded area highlights the periods under the curve for each negative period

Fig. 2

Heritability of Production and Resilience indicators. At the top of each bar we reported the symbol of significance of heritability associations (‘***’ for p-values lower than 0.001, ‘**’ for p-values lower than 0.01, ‘*’ for p-values lower than 0.05, ‘.’ for p-values lower than 0.1 Error bar represented confidence intervals mean ± 1.96 standard deviations

Fig. 3

Genetic (A) and phenotypic (B) correlations. Within cells, the number above represents the median of the correlation, and the number below represents the 5-95% interval of the empirical distribution (CI)

Fig. 4

(A) The Miami plot of BALnVar depicts the logarithm of residual variance with out adjustment in the upper part, with AFCD (Average Daily feed consumption) shown on the bottom. (B) The Miami plot of BAMaxArea illustrates the area under the curve for periods with the largest consecutive negative residual terms with Spline withs adjustment, with AFCD (Average Daily feed consumption) presented on the bottom. Dotted lines on both plots represent the suggestive and significant thresholds. The red dots signify the SNPs declared as “lead” by the fine mapping algorithm. Labels denote gene names within an interval of ± 0.1 Mb distance from the lead SNPs. Note that resilience traits were colored in blue while AFCD in orange

Fig. 5

Zoom plot of the significant region (157684006–163315197) on chromosome 1 associated with BALnVar logarithm of residual variance without adjustment. The upper part of the plot represents a zoomed Manhattan plot of the region described above. The color of SNPs represents the linkage disequilibrium of the SNPs within that region with the one declared as the “lead” by the fine mapping algorithm. The bottom part represents the physical position of genes within this region

Fig. 6

(A) The Miami plot of SPLnVar depicts the logarithm of residual variance with Spline adjustment in the upper part, with AFCD (Average Daily feed consumption) shown on the bottom. (B) The Miami plot of SPMaxArea illustrates the area under the curve for periods with the largest consecutive negative residual terms with Spline adjustment, with AFCD (Average Daily feed consumption) presented on the bottom. Dotted lines on both plots represent the suggestive and significant thresholds. The red dots signify the SNPs declared as “lead” by the fine mapping algorithm. Labels denote gene names within an interval of ± 0.1 Mb distance from the lead SNPs. Note that resilience traits were colored in blue while AFCD in orange

Fig. 7

Zoom plot of the significant region (23945248–25945215) on chromosome 4 associated with SPLnVar logarithm of residual variance with spline adjustment. The upper part of the plot represents a zoomed Manhattan plot of the region described above. The color of SNPs represents the linkage disequilibrium of the SNPs within that region with the one declared as the “lead” by the fine mapping algorithm. The bottom part represents the physical position of genes within this region

Fig. 8

Scatter plot of GO overrepresentation analysis for BALnVar and BAMaxArea. BALnVar depicts the logarithm of residual variance with out adjustment in the upper part, while BAMaxArea illustrates the area under the curve for periods with the largest consecutive negative residual terms with out adjustment

Fig. 9

Scatter plot of GO overrepresentation analysis for SPLnVar and BAMaxArea. SPLnVar depicts the logarithm of residual variance with Spline adjustment in the upper part, while SPMaxArea illustrates the area under the curve for periods with the largest consecutive negative residual terms