The Thoroughbred horse population is one of the largest closed populations of animals in the world. Thoroughbreds are extremely valuable because of the large amount of prizemoney on offer and the high residual value of superior athletes. All Thoroughbred horses trace their ancestry back to three paternal lines, due to the narrow bottleneck at the foundation of the population^1,2,3. More than 300 years of breeding practices have produced signatures of selection in the 21^st century Thoroughbred population, contributing to the superior athleticism of the breed^4,5. At the same time, these practices have increased levels of inbreeding and reduced the genetic diversity of Thoroughbreds compared with other domestic horse breeds^3,6,7.

To our knowledge, there has been no detailed examination of the effects of inbreeding on the racing performance of Thoroughbred horses and the genetic load of the population. Genetic load, the presence of unfavourable genetic material, is a reflection of a population’s fitness because a higher genetic load leads to a lower mean fitness level⁸. A large proportion of genetic load consists of recessive deleterious mutations, known as mutational load. Inbreeding can expose mutational load because it increases an individual’s chance of inheriting two copies of recessive deleterious alleles from a common ancestor^8,9. The subsequent decrease in fitness caused by these expressed recessive deleterious mutations is thought to be a major cause of inbreeding depression¹⁰. Other mechanisms believed to contribute to inbreeding depression include epistatic interactions and reductions in favourable heterozygosity^10,11.

The inevitable effect of selection in a closed population is an increase in the level of inbreeding^12,13. There is some evidence that continued inbreeding for selection can purge a population of some or all of its genetic load, such that new inbreeding events have negligible or even positive effects on phenotype⁹. Although some domestic and wild populations show signs of purging^14,15,16, others still show strong signs of inbreeding depression even after multiple population bottlenecks and inbreeding events^17,18,19. Purging is most likely to occur in populations under strong selection and slow rates of inbreeding, allowing deleterious alleles to be effectively eliminated rather than fixed by genetic drift^11,20. Additionally, inbreeding for favourable phenotypic characteristics can have unexpected negative implications through deleterious alleles hitchhiking on regions of the genome under positive selection, thereby increasing their frequency in the population^21,22,23.

Understanding the effects of selection is further complicated by the uneven distribution of genetic load in a population. Inbreeding to different ancestors can have varying effects on fitness, such that the total proportion of alleles identical by descent (IBD) might not be an accurate reflection of mutational load^24,25,26. This raises the possibility that inbreeding in different pedigree lines has variable effects on genetic load in the Thoroughbred population.

The availability of extensive phenotypic and pedigree records, dating back to the late 18^th century, makes the Thoroughbred population ideal for studying the long-term, population-wide effects of selection on performance and genetic load. Here, we examine the effects of inbreeding on racing performance and mutational load in the Australian Thoroughbred population. Australia has the second-largest racing and breeding population in the world, containing approximately 15% of all Thoroughbreds²⁷.

We analyse a sample of 135,572 individuals, representing all Thoroughbred horses that had one or more race starts in Australia between 2000 and 2011. A genealogy of these individuals, dating back to the founders of the population (n = 257,249), is also included in our analyses. Although some lines of pedigree are incomplete, we have comprehensive pedigree information for all individuals in the racing performance data set, making our inbreeding estimates highly accurate. The availability of extensive pedigree records not only allows us to study broad population trends over time, but also to determine whether the selection for optimal racing performance has alleviated mutational load. We use these data to measure inbreeding and ancestral coefficients for all individuals. We also identify the ancestors that have made the greatest genetic contributions, in order to understand better the distribution of mutational load in the population. For a representative subset of individuals, we perform high-density genotyping to determine whether inbreeding load is reflected at the genomic level.