Update by Lady Fi, 2016: see my recent article as a followup to this one. Researchers confirm PRA test dodgy
Any breeder who has ever modified their breeding program based on DNA test results alone will find this post concerning.
Right across the world dog breeders are turning to DNA testing as an important breeding management tool. Indeed, some dog associations have made DNA screening of certain breeds and diseases obligatory.
Going one step further, in the State of Victoria, Australia, the breeding of dogs with heritable diseases was recently written into criminal law.
Before I expound further, I wish to give context by making clear my position here. My position is that I one hundred percent support the rigorous screening of breeding dogs for heritable diseases. I am one hundred percent behind us better understanding the way diseases are passed through the generations and for the development of predictive tests that breeders can use in their pursuit of breeding high health dogs.
Indeed, it is this very commitment to breeding genetically sound dogs that underpins this article and my earlier one. Namely, if a test (be it DNA or other method) is not interpreted correctly, we could be throwing out the baby with the bathwater. That is, we could be losing lots of great genes (not just the supposed ‘bad ones’) as we misguidedly modify our breeding programs to steadily breed away from certain genes in the erroneous belief that they are deleterious.
Now I will continue.
I have just recently met with scientists from a genetic testing company who were helpful in explaining ‘how it all works’ from both a methodology and technical viewpoint. I found them completely upfront in their explanation and they showed me their own brochures where the main issue of what I am about to address was spelt out (that is, lack of evidence between disease test results and disease occurence). Other genetic testing company websites have similar explanations concerning the low or unknown link between a positive disease test result and the actual disease.
Somewhere though, it is fair to say, the ‘how it all works’ bit has been lost in the shift from laboratory to industry application, as evidenced by the fact that people throughout the world are using DNA test results alone to modify breeding programs.
How does it work?
DNA disease tests do not test for whether or not an animal will get any given disease. Rather, they test for the presence of certain genes that have been found to occur in animals that actually physically have the disease. Or, putting it in science-speak, the test is a description of one aspect of the animal’s genotype (its genetic make-up), but not its all-important phenotype (what it appears like in real life, including whether or not it will get a disease one day).
As far as predictive tests go, this is not unusual in itself because by definition predictive tests must be based on identifying some factor that is a pointer to disease onset. Smoking and lung cancer is a good example: if you smoke you increase your chances of getting lung cancer by X% for a given demographic.
The problem with DNA testing and dogs, however, is that the all-important ‘correlation step’ (the ‘X% bit’ in the smoking/cancer example) is absent for many of the dog DNA disease tests. That is, no link is made between the likelihood that dogs having the genes will actually develop the disease. This leaves me scratching my head thinking, ‘what possible value can any test have if the ‘something’ it identifies has an unknown – possibly zero – relationship with the disease it purports to predict?‘ Hardly a predictive pointer, is it?
I explored this seemingly gaping hole in scientific process with the genetic testing company scientists. Specifically, I asked what proportion of animals tested as ‘Affected’ (the term universally used to describe animals carrying two copies of a given gene) actually develop the disease at any stage in their lives. That is, how many DNA tested animals labelled ‘Affected‘ actually turn out to be affected by the disease?
Their answer was that it was not something they had data on (nor do other DNA companies, from my understanding) and that they would like this data, but it is hard to get from vets. Rather, they said, it was up to the breeder concerned to decide the likelihood of any gene-disease link based on their knowledge of whether the disease is a problem in their breed.
Pursuing this point further, I brought up my example of PRA blindness in miniature smooth dachshunds as a case in point. Namely, the UK Kennel Club has screened 1000 plus miniature smooth dachshunds and, I presume, is modifying breeding recommendations directly based on these results alone. The Club states that an ‘affected’ test result means a dog has an ‘increased risk’ of developing PRA in its lifetime. To this, I say: what evidence is there to support this statment? What evidence is there to show that the incidence of ‘Affected’ mini smooth dachshsunds contracting PRA blindness is any higher than the rest of the dog population?
I explained to the DNA company scientists how my research to date has not found there to be a correlation between the DNA PRA disease test and the PRA disease being a problem in real life for mini smooth dachshunds.
In a case such as this, I was told, a breeder should obviously ignore the DNA PRA test result because it is not a good indicator of disease occurrence in real life.
Golly gosh. Tell that to 1000 smooth mini dachshund breeders in the UK Kennel Club. Tell that to the breeders in Victoria, Australia, now adopting a government sanctioned ‘recognised breeding program’ for PRA under threat of criminal prosecution should they do any different. And tell that to the owner of a much adored fine dog now publicly labelled ‘Affected’.
If this lack of correlation between genes and disease exists in the case of the mini smooth dachshund and PRA, how many other breeds are in a similar situation with this or other DNA disease tests? For the same scientific approach and process holds for all. Namely, the correlation between genes and disease occurrence is either not explored or exploration is ad hoc and limited.
How do they find these genes?
It is my understanding that genes are implicated as disease causing when scientists study a sample population of diseased animals and find them all to possess some common gene. That common gene is therefore tagged the culprit.
When this happens, scientists can accurately and confidently state that every animal exhibiting the disease, in the sample population examined, possesses the identified gene.
This logic might sound fine: who needs more proof to link cause and effect? It sounds fine until, that is, I give you an analogy:
All rabbits have two ears, therefore all creatures with two ears
Rewriting the rabbit analogy for PRA and mini dachshunds:
All diseased miniature dachshunds with PRA blindness possess a certain gene, therefore all dachshunds with said-gene will too have this disease.
Why such misinterpretation?
I believe I have found an explanation as to why universal misinterpretation of DNA test results persists. Namely: miscommunication and misleading terminology.
This all begins with the fact that in certain cases it is true that DNA tests are all-predictive and thereby all-conclusive. That is, there are times when all two eared creatures are rabbits – in a heritable context. This occurs in cases where simple genetic Mendelian laws of ‘recessive and dominant single-gene’ inheritance are at play.
There are some characteristics of dogs that do follow such simple inheritance laws. Coat type comes to mind. I regularly cross my long and smooth coated dachshunds with the result being all smooth coated pups. This is because there is only one gene associated with each coat type, and there is a hierarchy of dominance and recessiveness at play.
In the case of coat type, smooth dominates over long hair dominates over wirehaired. If I DNA test my dogs for the long haired gene, therefore, the results will be accurately predictive. Thanks to Mendel, this correlation between gene and outcome is ‘a given’. No further research needed.
In such an example of simple inheritance, where one gene alone codes for a characteristic, we can be confident of breeding outcomes with no further research required. The correlation between the predictor (the gene) and the outcome (coat type, in this example) is one and the same.
And here is where I see the first slip up in communication or translation to be taking place as we shift from scientist to breeding practitioner.
Perhaps in the desire to keep things ‘simple’, heritable diseases that do not operate on the simple Mendelian laws of inheritance (i.e. the monogenic recessive vs dominant gene theory) have been incorrectly described as thus.
For example, harking back to the PRA situation we see on the UK Kennel Club website that the explanation accompanying the publishing of 1000 plus dog names and their parents is that a dog can be one of only three outcomes:
- Clear – has no PRA gene
- Carrier – has one PRA gene
- Affected – has two PRA genes
Herein the error has begun, for this preamble implies that simple ‘recessive-dominant’ inheritance applies, whereas the facts (of negligible PRA blindness in smooth mini dachshunds) tell us differently.
What should be stated instead, and I believe DNA test company industry colleagues would agree here, is something like:
‘There is no data currently linking this gene with the incidence of the disease PRA in miniature smooth dachshunds occurring at a frequency above that in the global dog population. Further, based on current anecdotal information and in the absence of any scientific data, the correlation appears poor.’
The facts tell us that in the case of the disease PRA blindness in smooth haired miniature dachshunds there is no simple single-gene action operating, but rather more complex gene interactions at play. Yet the simplest model of inheritance has for some unknown reason been applied in this case; in words, action and law.
I stand to be corrected by a geneticist, but it is my understanding that the majority of traits in animals are not coded for by simple monogenic inheritance, but rather multi gene complexes. This is further complicated by the fact that some genetic traits need specific environmental cues to trigger them to surface. In other words, the majority of traits (with diseases among them) are tied in to many factors and these can be both within the animal and external to the animal.
In regard to misleading terminology, it is unfortunate that the technical name used to describe an animal with two copies of the gene of interest is ‘Affected’. For in reality we now know that many of these dogs will be anything but ‘affected’ when it comes to ever getting the disease. Nonetheless, ‘Affected’ is the name they are tagged with along with all its implications.
The reason this word ‘Affected’ is used to identify hymozygous animals (those with two gene copies) relates back to the use of simple Mendelian laws being applied in cases where clearly more complex genetics is at work, as discussed earlier. For under the simple monogenic inheritance, possession of two genes would indeed make the animal ‘affected’. Two recessive genes for ‘long coat’ make my dachshund have a long coat: perfect correlation in a simple inheritance scenario and hence a logical explanation for the use of this word. Transfer the word to complex gene inheritance scenarios, however, and it becomes misleading.
Unfortunately, the real world of animal breeding isn’t all simple: not all animals with two ears are rabbits.
If we persist in interpreting DNA disease test results incorrectly and imposing criminal laws and club rules accordingly, we undermine the health and longevity of the very animals whose welfare we would purport to protect.
Feel free to comment publicly here. You can do so anonymously. Breeders, legislators, genetic testing companies, researchers, dog owners, dog associations, dog lovers: your insight is welcome and needed.