It is hoped that the public availability of population analysis reports for all 215 breeds currently recognised by the Kennel Club will enable breeders and other stakeholders to achieve a better understanding of the unique situation facing each breed. Both the descriptors of population dynamics (i.e. number of registrations, number of sires used etc.) and the genetic parameters (i.e. ΔF and N_e) inform of the recent breed history and of the necessary considerations framing potential breeding strategies in the future. Different breeds with varying dynamics and levels of genetic diversity will have different options available in balancing meeting selection objectives and maintaining a sustainable ΔF. The breed reports are intended to provide a backdrop to discussions over the best approaches.

The plot of observed inbreeding for Labrador Retrievers (Fig. 2) is fairly typical of the profile of most breeds, with a steep incline in the 1980s gradually flattening to present, and in some breeds even recently declining. This corresponds to the general trend detected across all breeds (with sufficient numbers registered) of a high ΔF in the 1980s and 1990s, followed by a general decline to a negative ΔF latterly (Fig. 4, where the magnitude of the bars relates to the gradient of ‘observed inbreeding’ in Fig. 2). The across breeds mean ΔF in the blocks 1980–1984, 1985–1989, 1990–1994 and 1995–1999 all far exceed the recommended maximum of 0.01 per generation, above which the detrimental consequences of inbreeding are expected to be observed (i.e. N_e < 50; [11]). Thus, general breeding practices in the 1980s and 1990s appear to have resulted in a major contraction of within breed genetic diversity, while practices latterly have, to some extent, ameliorated this. The UK quarantine laws requiring immigrant dogs spend a six month period at quarantine kennels were relaxed in 2000 with the introduction of the Pet Passport scheme. It may be expected that the number of dogs imported to the UK after this date would have increased, and the use of apparently more distantly related migrant animals for breeding would contribute to the recent decline in ΔF. Although broadly the same declining trend in ΔF was observed in numerically small and large breeds (Fig. 6 and 7 respectively), the extent of the change was greater in the VNBs. It may be that the effects of general breeding practices were exacerbated in numerically small breeds (VNBs) in earlier years and that the increased availability of breeding stock from outside the UK from 2000 onwards provided a much needed injection of genetic diversity. In the case of the consistently more common breeds, the smaller mean ΔF in early years may reflect a larger pool of potential breeding animals, and the slower decline in mean ΔF due to a smaller proportional impact of migrants.

The whole period N_e varied widely across breeds, ranging from 23.8 to 918.8 (where determinable, in breeds with an average of >50 registrations per annum over each 5-year block). Individual breed whole period N_e represents the slope of ‘best fit’ through the ‘observed inbreeding’ plot (example in Fig. 2) over 1980–2014, describing the idealised population size that would be expected to exhibit the ΔF observed over 35 years. As such, it does not account for fluctuations in ΔF within that 35 year period, which have been shown in the results and discussed above. It is therefore important to take into account the individual breed profile in ‘observed inbreeding’ and ΔF and N_e over the 5-year blocks to determine the extent of contraction in genetic diversity, when it occurred, and the degree to which this may have been restored more recently to inform future breeding strategies on a breed-by-breed basis.

The whole period N_e was independent of census population size (as judged by mean annual registrations); some very numerous breeds had a small whole period N_e (e.g. English Springer Spaniel, mean annual registrations = 10,885.7, N_e = 45) while some much rarer breeds had a relatively high whole period N_e (e.g. Sealyham Terrier, mean annual registrations = 87.1, N_e = 111). A possible reason for low N_e in numerically large breeds is underlying population ‘sub-structure’. In common breeds, such as the English Springer Spaniel, several ‘sub-populations’ are likely to exist; for example working, show and pet populations, and even geographically localised populations. The existence of sub-populations (to whatever degree of independence) is indeterminable in this analysis of pedigree data. Breeding practices within each of multiple independent sub-populations giving rise to a positive ΔF in each will lead to a positive breed-wide ΔF. However, this breed-wide figure represents [mean] loss of genetic diversity within each sub-population, but ignores the fact that there is actually an increase in genetic diversity between sub-populations due to drift acting on allele frequencies [9]. Thus the breed-wide estimate of ΔF and N_e fails to take account of the between sub-population genetic variation, which is easily tapped by migration between sub-populations. Therefore, for some of the numerically larger breeds with show and working ‘types’ , the breed-wide whole period N_e may belie the amount of true within breed genetic diversity.

Although the lack of complete pedigree data beyond a certain number of generations for migrant animals hampers the ability to fully determine co-ancestry where it does exist (therefore potentially underestimating ΔF and N_e for breeds making wide use of imported animals), the principle outlined above also applies to breed populations in different countries. Even where such migrant animals originally trace back to the UK (for example the English Setter), populations which have existed in semi-isolation in different countries will be subject to the effects of drift, potentially increasing genetic diversity between sub-populations while it simultaneously declines within each. However, to what degree the increase in between sub-population diversity counteracts the potential underestimation of ΔF and N_e due to incomplete pedigree information is unknown.

The sustainable ΔF and N_e observed in some numerically smaller breeds (e.g. Dandie Dinmont Terrier, Cardigan Welsh Corgi, Sealyham Terrier, Bloodhound) would appear to be related to the effective management of genetic diversity. The plots of observed and expected inbreeding in these breeds show small divergence between the two, implying only a slight departure from random mating (Additional file 4: Figure S1). This may be due to heightened breeder awareness of the importance of conserving genetic diversity in numerically small populations. However, in some cases where the breed is numerically small, effective management of genetic diversity may not be enough. Otterhound breeders appear to have been managing genetic diversity as effectively as possible over 1980–2014 (judged by the conformity of the plots of observed and expected inbreeding), and so the high ΔF and low N_e observed may be due to small actual population size (Additional file 5: Figure S2). There are methods which achieve lower ΔF than predicted via random mating which may be useful in the preservation of genetic diversity, for example negative assortative mating and optimum selection techniques where weighting is only placed on minimising ΔF [12, 13]. However, these methods rely on tightly co-ordinated decisions being made on breeding animals, which is unlikely even in the rarest of breeds.

Furthermore, preserving genetic diversity (via a sustainable ΔF) is a single, albeit important, objective among many in the promotion of health and welfare in dog breeds. Many breeds face significant welfare problems due to a high burden of inherited disease [4] and extreme conformation [14], for which a lasting and widespread improvement can only be achieved via selection. Selection requires genetic variation to be present within the population meaning that the desired response will be more difficult to achieve in populations with a high ΔF. Partial restoration of genetic variation will occur via mutation, but the rate is of mutation is small, the genomic location random and there is strong selection pressure against mutations in coding regions of the genome, where mutations lead to malfunction or loss of function. Therefore, where genetic variation is depleted in a population, migration of breeding animals from a different population is the only practical means of regeneration to enable selection. Selection and inbreeding have been demonstrated to be fundamentally related via genetic contributions [6–8], meaning that balancing the competing objectives of genetic gain and sustainable ΔF is vital to the welfare and sustainability of many breeds. The intrinsic relationship between selection and inbreeding means that in numerically larger breeds the ΔF observed in these analyses could be considered to be a ‘signature’ of selection, although the objectives and traits under selection remain unknown. Therefore, breeds with very low N_e despite moderate to high mean registrations per year (e.g. Airedale Terrier, Bearded Collie, Irish Setter, Yorkshire Terrier; Additional file 6: Figure S3) might be considered to have been subject to relatively intense selection over the past 35 years.

While the use of inbreeding coefficients to derive ΔF in a breed or population provides a useful indication of past practice, they remain a retrospective measure of co-ancestry. Well-intentioned but widespread use of unrelated animals in breeding programmes in an attempt to increase genetic diversity, while reducing mean F in the next generation, can lead to such individuals becoming popular sires. Popular sires make a large genetic contribution to future generations of the breed, and are therefore the major contributor to a high ΔF over subsequent generations [15]. Breeders of pedigree breeds making particular use of migrant animals must be aware of this. Using a greater proportion of males for breeding will help mitigate the effect popular sires have on ΔF, but the common practice of substituting a known popular sire with a close male relative in a potential mating will have limited impact on minimising rises in future ΔF, due to shared genetics. The monitoring of genetic contributions may allow prospective identification of potential over-popular sires and relatives, and research into how such a strategy may be tailored to dog breeding is ongoing [16].