How Depressing Is Inbreeding? A Meta-Analysis of 30 Years of Research on the Effects of Inbreeding in Livestock

Abstract

Inbreeding depression has been widely documented for livestock and other animal and plant populations. Inbreeding is generally expected to have a stronger unfavorable effect on fitness traits than on other traits. Traditionally, the degree of inbreeding depression in livestock has been estimated as the slope of the linear regression of phenotypic values on pedigree-based inbreeding coefficients. With the increasing availability of SNP-data, pedigree inbreeding can now be replaced by SNP-based measures. We performed a meta-analysis of 154 studies, published from 1990 to 2020 on seven livestock species, and compared the degree of inbreeding depression (1) across different trait groups, and (2) across different pedigree-based and SNP-based measures of inbreeding. Across all studies and traits, a 1% increase in pedigree inbreeding was associated with a median decrease in phenotypic value of 0.13% of a trait's mean, or 0.59% of a trait's standard deviation. Inbreeding had an unfavorable effect on all sorts of traits and there was no evidence for a stronger effect on primary fitness traits (e.g., reproduction/survival traits) than on other traits (e.g., production traits or morphological traits). p-values of inbreeding depression estimates were smaller for SNP-based inbreeding measures than for pedigree inbreeding, suggesting more power for SNP-based measures. There were no consistent differences in p-values for percentage of homozygous SNPs, inbreeding based on runs of homozygosity (ROH) or inbreeding based on a genomic relationship matrix. The number of studies that directly compares these different measures, however, is limited and comparisons are furthermore complicated by differences in scale and arbitrary definitions of particularly ROH-based inbreeding. To facilitate comparisons across studies in future, we provide the dataset with inbreeding depression estimates of 154 studies and stress the importance of always reporting detailed information (on traits, inbreeding coefficients, and models used) along with inbreeding depression estimates.

Keywords: cattle; chicken; cow; goat; homozygosity; horse; pig; rabbit; sheep.

Figure 1

Histograms of estimates of $b_m$ (n = 1818) and $b_s$ (n = 1259) across all studies and traits, after removal of extreme outliers. Descriptive statistics and a normal distribution (dashed red lines; based on mean and standard deviation (SD)) are also shown.

Figure 2

Violin plots of inbreeding depression estimates per trait group. Estimates are expressed as a percentage of a trait’s mean ($b_m$) or as a percentage of a trait’s SD ($b_s$). Boxplots are also shown, indicating the median, 25th and 75th quantiles and the mean ($\times$) for each group. For $b_m$ and $b_s$, there were respectively 40 and 39 extreme estimates outside the range of this figure.

Figure 3

Relationship between inbreeding depression estimates expressed as percentage of a trait’s mean per 1 standard deviation increase in inbreeding ($b_m$ * SD(F)) across different measures of inbreeding. The data points (colored per study) and linear trendline are shown (lower triangle) as well as the density curve for each inbreeding measure (diagonal) and the correlation and regression equation (upper triangle). Note that slopes of the linear trendline differ from 1, which is also expected when correlations between inbreeding measures themselves are not equal to 1. $F_{PED}$ = pedigree inbreeding; $F_{ROH}$ = inbreeding based on runs of homozygosity; $F_{GRM}$ = inbreeding from genomic relationship matrix (studies in pink and purple used VanRaden’s method 2, and light blue Yang’s method); $F_{GRM0.5}$ = inbreeding from genomic relationship matrix with allele frequencies fixed to 0.5; $HOM$ = percentage of homozygous SNPs.

Figure 4

Relationship between inbreeding depression estimates expressed as percentage of a trait’s standard deviation per 1 standard deviation increase in inbreeding ($b_s$ * SD(F)) across different measures of inbreeding. The data points (colored per study) and linear trendline are shown (lower triangle), as well as the density curve for each inbreeding measure (diagonal) and the correlation and regression equation (upper triangle). Note that slopes of the linear trendline differ from 1, which is also expected when correlations between inbreeding measures themselves are not equal to 1. $F_{PED}$ = pedigree inbreeding; $F_{ROH}$ = inbreeding based on runs of homozygosity; $F_{GRM}$ = inbreeding from genomic relationship matrix (studies in pink and purple used VanRaden’s method 2, and light blue used Yang’s method); $F_{GRM0.5}$ = inbreeding from genomic relationship matrix with allele frequencies fixed to 0.5; $HOM$ = percentage of homozygous SNPs.

Figure 5

Example with three normal distributions (left; each with a mean of −0.22 and a SD of 0.2, 0.6 or 1) and the resulting mixture of these three normal distributions (right), showing an increase in kurtosis.

Figure 6

Funnel plot to assess publication bias. The plot shows the relationship between inbreeding depression estimates, expressed as a percentage of the trait mean ($b_m$), and the number of records used to estimate them (N = 1283). The orange vertical line represents the median. To ease interpretation, estimates based on >400,000 records are not shown (N = 82).