DogsArk
Golden Retriever
GENETIC DASHBOARD
SYNOPSIS
Genomic inbreeding (F) in Golden Retrievers is very high, averaging 23.5%, with many dogs above 30% and some > 40%. On the other hand, inbreeding of some dogs in the population is less than 15% and even as low as 1.7%. The fixation index (Fis) for most dogs is positive, reflecting selection of breeding pairs that are more closely related than the average for the population (i.e., Fis > 0). (You would expect mean Fis = 0 in a randomly breeding population.) Despite high levels of inbreeding, the average mean kinship (mK) was only 4.5% (range 0.4% to 11.3%), indicating that selecting dogs as parents with low mK would produce litters with lower inbreeding than in this population. Inbreeding is based on the proportion of homozygous loci that results from replacement of one allele of a heterozygous locus with a pair that are identical by descent (IBD; inherited from an ancestor common to both sire and dam). This reduces the diversity of functional alleles and increases the risk of producing loci that are homozygous for a mutation. Where these homozygous loci are recessive mutations, there will be loss of function.
The high level of inbreeding in this population of Golden Retrievers indicates that we should expect to see significant compromise in the functions and processes critical for health, growth, and reproduction. However, there might be significant useful diversity in some other populations of dogs (eg., hunting vs bench, other countries, etc), and it is definitely worth collecting the necessary samples to characterize this potential variation.
The data presented here are summarized on DogsArk.
SUMMARY
The Dogs and Data
This analysis uses DNA genotyping data obtained from Adam Boyko (Cornell, Embark) that was used in a study published in 2016. This included data for 247 Golden Retrievers. The identity or origin of these dogs is not known as the data were anonymous, but most were probably from the US. There is no health information about these dogs; simply an ID and genotype date. Note that because of the date of the study (2016), the file is unlikely to include any currently breeding dogs.
I processed the data using standard protocols for SNP analysis on the Illumina HD (high density) Canine Bead Chip. The analyses were performed using Golden Helix SNP and Variation Suite software. The algorithms the software uses are sourced to published studies validating their output.
Inbreeding (F)
The graph below displays four measures of inbreeding and diversity: genomic inbreeding, fixation index, kinship, and observed heterozygosity.
Inbreeding (F) averaged 0.235 (23.5%), in this population which is about the level of inbreeding (25%) you would expect from mating of two siblings from unknown parents. Values of F ranged from 1.7% to 46.7%.
Fixation Index (Fis)
The inbreeding of each dog relative to the average for the population is the “fixation index” (Fis). These values will be negative if the inbreeding of the dog is less than the population average, or positive if the dog is more inbred than the average. A dog with inbreeding the same as the population average will has Fis = 0.
The data for Fis reveal that breeders are producing dogs with high inbreeding by preferentially pairing dogs that are closely related (Fis > 0).
Observed Heterozygosity (Ho)
Heterozygosity is a measure of the amount of genetic variation in a population. Inbreeding, selection, and genetic drift can result in the loss of alleles from the gene pool, which will be reflected in lower heterozygosity. If every locus is heterozygous, the observed heterozygosity (Ho) will be 0.5.
The data for observed heterozygosity (Ho) indicate substantial loss of genetic diversity in Golden Retrievers. This is consistent with high inbreeding and also probably reflects low diversity in the foundation dogs.
Kinship (K)
The level of inbreeding in a litter of puppies is predicted by the degree of relatedness or genetic similarity of the parents. This is quantified by the kinship coefficient (K); that is, the kinship coefficient of a bitch and sire is equal to the predicted (average) inbreeding of their hypothetical litter.
The mean kinship (mK) for a dog is the average of all of the potential pair-wise kinship coefficients in the population of interest. Dogs that are closely related to many others in the population (e.g., offspring of a popular sire) will have a high mK; dogs carrying alleles that are uncommon in the population will be genetically less similar on average to the rest of the population and, for them, mK will be relatively low.
The histogram of mean kinship for the animals in this population is skewed towards lower values, i.e., < 0.12. This is consistent with the data for Fis, which together show that breeders are avoiding pairings of animals that are less related than average (i.e., preferentially choosing closely related parents). If breeding was random in this population, the data for inbreeding and mean kinship would have similar distributions.
These data reveal that most dogs could be paired with a mate to produce litters with F < 12%. Judicious use of the existing genetic diversity in this population could reduce inbreeding to the low single digits. Breeders could easily identify these potential pairs of parents using kinship coefficients in the form of a matrix that displays in a single chart the K for all pairings of potential interest.
I did not create a kinship matrix for Goldens because of the large number of dogs. It would be more useful to do after identifying specific dogs of interest.
The dogs in a population that carry alleles that are less common in the breed have the highest genetic value. Conversely, dogs with high genetic similarity to many other dogs are least valuable; the loss of one of these dogs would have little impact on the composition of the gene pool of the breed. To protect the genetic diversity of the population, the dogs with the highest genetic value should be used in breeding.
The easiest way to identify the dogs with the highest genetic value is by ranking them by mean kinship; the dogs with the lowest mean kinship have the highest genetic value. The most valuable dogs in this population are at the left with the lowest mean kinship values. (The IDs of the dogs are beneath each bar but too small to read here.)
The inbreeding of each dog relative to the average for the population is the “fixation index” (Fis). These values will be negative if the inbreeding of the dog is less than the population average, or positive if the dog is more inbred than the average. A dog with inbreeding the same as the population average will has Fis = 0.
The data for Fis reveal that breeders are producing dogs with high inbreeding by preferentially pairing dogs that are closely related (Fis > 0).
Observed Heterozygosity (Ho)
Heterozygosity is a measure of the amount of genetic variation in a population. Inbreeding, selection, and genetic drift can result in the loss of alleles from the gene pool, which will be reflected in lower heterozygosity. If every locus is heterozygous, the observed heterozygosity (Ho) will be 0.5.
The data for observed heterozygosity (Ho) indicate substantial loss of genetic diversity in Golden Retrievers. This is consistent with high inbreeding and also probably reflects low diversity in the foundation dogs.
Kinship (K)
The level of inbreeding in a litter of puppies is predicted by the degree of relatedness or genetic similarity of the parents. This is quantified by the kinship coefficient (K); that is, the kinship coefficient of a bitch and sire is equal to the predicted (average) inbreeding of their hypothetical litter.
The mean kinship (mK) for a dog is the average of all of the potential pair-wise kinship coefficients in the population of interest. Dogs that are closely related to many others in the population (e.g., offspring of a popular sire) will have a high mK; dogs carrying alleles that are uncommon in the population will be genetically less similar on average to the rest of the population and, for them, mK will be relatively low.
The histogram of mean kinship for the animals in this population is skewed towards lower values, i.e., < 0.12. This is consistent with the data for Fis, which together show that breeders are avoiding pairings of animals that are less related than average (i.e., preferentially choosing closely related parents). If breeding was random in this population, the data for inbreeding and mean kinship would have similar distributions.
These data reveal that most dogs could be paired with a mate to produce litters with F < 12%. Judicious use of the existing genetic diversity in this population could reduce inbreeding to the low single digits. Breeders could easily identify these potential pairs of parents using kinship coefficients in the form of a matrix that displays in a single chart the K for all pairings of potential interest.
I did not create a kinship matrix for Goldens because of the large number of dogs. It would be more useful to do after identifying specific dogs of interest.
The dogs in a population that carry alleles that are less common in the breed have the highest genetic value. Conversely, dogs with high genetic similarity to many other dogs are least valuable; the loss of one of these dogs would have little impact on the composition of the gene pool of the breed. To protect the genetic diversity of the population, the dogs with the highest genetic value should be used in breeding.
The easiest way to identify the dogs with the highest genetic value is by ranking them by mean kinship; the dogs with the lowest mean kinship have the highest genetic value. The most valuable dogs in this population are at the left with the lowest mean kinship values. (The IDs of the dogs are beneath each bar but too small to read here.)
Assessing the Risk of Genetic Disorders Without Genes
Kinship coefficients can be used to identify dogs potentially at risk for certain disorders, even when the genes responsible are not known. (See my post about how to find trait or disease genes without DNA.)
Cluster analysis of the kinship data is used to identify groups dogs that are genetically similar. This information is displayed as a dendrogram (tree of relationships) and individuals with the traits of interest are identified. Patterns in occurrence within and among indicate lineages higher risk than others. This allows breeders to select for or against traits for which the genetic basis is unknown, even for polygenic traits or disorders.
I do not have health information for this population of Golden Retrievers, but this valuable analysis can be done if genotype and health data are provided to DogsArk for Goldens. (See the analyses for other breeds for examples)
For more information about using kinship coefficients in dendrograms to evaluate the potential expression of heritable traits see Cool tricks with Kinship Coefficients, part 1: "Is this dog really an outcross?".
For more information about using cluster analysis and dendrograms to explore genetic patterns in disease and traits in a breed, see “Cool tricks with kinship coefficients, part 3: “How can I manage a disease without a DNA test?”
Finding Genetic Diversity in a Breed
If breeders wish to improve the genetic diversity of their dogs through an outcross, how can they identify dogs that are genetically different? One way is by examining the dendrogram that results from cluster analysis of kinship coefficients.
There is another tool called “principal components analysis”, which is a statistical technique that clusters together genetically similar dogs. Instead of a dendrogram, the analysis produces a scatter plot on two-dimensional axes. The genetic similarity between individuals is shown by the distance between them. So, a population of animals might be depicted as multiple subpopulations identified as clusters of dogs. In the PCA plot for the Golden Retrievers in this sample, most dogs fall in a large cluster across the middle of the space, but there are some dogs that fall in the lower left quadrant of the graph. These differences in distribution might identify field vs bench dogs, UK vs US dogs, or any other subgroupings that for one reason or another have become genetically less similar to other groups through selection, breeding isolation, or genetic drift
If breeders wish to improve the genetic diversity of their dogs through an outcross, how can they identify dogs that are genetically different? One way is by examining the dendrogram that results from cluster analysis of kinship coefficients.
There is another tool called “principal components analysis”, which is a statistical technique that clusters together genetically similar dogs. Instead of a dendrogram, the analysis produces a scatter plot on two-dimensional axes. The genetic similarity between individuals is shown by the distance between them. So, a population of animals might be depicted as multiple subpopulations identified as clusters of dogs. In the PCA plot for the Golden Retrievers in this sample, most dogs fall in a large cluster across the middle of the space, but there are some dogs that fall in the lower left quadrant of the graph. These differences in distribution might identify field vs bench dogs, UK vs US dogs, or any other subgroupings that for one reason or another have become genetically less similar to other groups through selection, breeding isolation, or genetic drift
Potential Impacts of Breeding Strategy on Health
This information indicates that breeders who wish to prioritize health could produce litters with lower inbreeding through choice of less related parents. Remember that the risk of producing genetic disorders from recessive mutations is proportional to homozygosity. Consequently, the most effective way to reduce the frequency of genetic disorders caused by recessive mutations is not through DNA testing, as commonly assumed, but by reducing inbreeding (F). DNA tests will not produce healthy dogs as long as there are other mutations lurking in the gene pool that we can’t test for. However, a reduction in inbreeding (reduction in homozygosity) will reduce the risk of ALL such recessive disorders. Reducing homozygosity should also reduce inbreeding depression, with positive effects on health, function, and lifespan. The negative effects of inbreeding Goldens have been documented for various fitness traits such as litter size and lifespan (Chu et al 2019; Yordy et al 2020; Kraus et al 2023).
