Eldin Leighton, PhD (pt. 2): The International Working Dog Registry

Jessica Hekman: Hi, friends. This week we’re back again with Eldin Leighton. After a long career at The Seeing Eye, Eldin founded the International Working Dog Registry, or IWDR. If you didn’t listen to last week’s episode where Eldin talked about his career before the IWDR, definitely check that out before listening to this one.

In this episode, he tells us all about the IWDR. How it helps gather large populations of dogs together to provide more data and analyze that data to help breeders improve their breeding decisions. And whether it’s useful for more than just working dogs. Spoiler. Yes, it is.

Eldin, thank you so much for coming back and making the time to talk to us again. It’s fabulous to get so much of your time. So I think that we had talked about estimated breeding values and how useful they are for breeding guide dogs successfully in our last talk. And so anybody who missed that, I strongly encourage you to go listen to that talk first, that interview first, before trying to tackle this one. But I guess my opening question for you today is, if estimated breeding values, which we’ll call EBVs, are so great than are all guide dog schools using them? Or, if some guide dog schools are not using them, why not?

Eldin Leighton: That’s a great question, Jessica. At this point in 2020, I can name three organizations for sure that are using estimated breeding values, and a fourth one that possibly is using them. And those four are four of the largest guide dog organizations in the world. There is not a service dog organization I am aware of which is using EBVs. And that raises the really interesting question, why not? And the reason is because it requires organizing data into a uniformly coded scheme, where records can be properly cleansed and made useful for feeding into the EBV calculation routines. And only when you have done all those things and laid the proper foundation and groundwork, are you able to actually calculate EBVs and make them readily available.

So what we’re doing at IWDR is providing a method for doing the proper coding of records as they’re entered, and maintaining those records in a way that ultimately makes it easier to begin using those data to calculate estimated breeding values. So that’s the real purpose of IWDR and the role it plays. And that’s why most organizations haven’t used it because they’re not large enough to create a database of their own. And even if they were large enough, there are only a few of these working dog organizations that produce enough puppies where the dynamics of population biology would enable the use of estimated breeding values to actually work.

JH: So you basically need a big population and the guide dog schools are split up into lots of little populations. And so that being a problem for them, then IWDR came along, which I don’t recall whether we talked… I’m not sure that we actually got to talk about that in our last episode. So let’s, and I’m hoping we can really focus on it for this episode. So you saw a need to bring these little populations together into one big population, basically.

4:48

EL: We did, and we were actually encouraged in that venture, or adventure, by a golden retriever breeder who desperately wanted to use these techniques to genetically improve dogs she produces. But there was absolutely nothing available. There are commercial databases written for use by dog breeders, but they all just focus on standard pedigrees and they do a basic calculation of inbreeding on every dog. But there’s nothing any place in the world which is a globally accessible tool for capturing the data in a way that it’s intended from the get go to calculate EBVs. So these small organizations all had a desire. We’ve been talking about and teaching all these organizations about estimated breeding value since at least 2006. And they, they wanted to use them, but there simply was no tool available. So we finally decided the only way such a tool was going to be possible was if we created it.

JH: And who was we?

EL: Jane Rusenberger and I are the two primary people. We also received a ton of help from a good friend and wizard at programming, Kevin Keymer. Unfortunately, he passed away two years ago, and so we have replaced his help with my son, Samuel. But Sam doesn’t have the experience of Kevin, so Sam is learning and is a wizard at things that Kevin was not. And so he’s taking on a new role and helping us do the programming, but we really miss Kevin as well. And at the same time, we’re extremely appreciative of having all of Sam’s help. So.

JH: Yeah, that’s a big loss.

EL: That’s the team. It’s a team of three and has been a team of three, basically from the get go.

JH: So when did this… When did you pull this team together?

EL: It happened gradually over time. Jane and I have tried to figure out exactly when we started really talking about it in earnest. We think it was 2008 or 2010.

JH: I didn’t realize it was that long ago you’d been talking about this?

EL: Could have been 2012. But yeah, we’ve been working on it seems like forever. (laughter)

JH: So, and my recollection is that you three basically started building it just as its own thing. And we’ll talk in a minute about how it’s now part of IWDBA, International Working Dog Breeders Association?

EL: Breeding Association.

JH: But at first you three just put it together yourselves. So what were your goals for it? What did you want it to be able to do?

EL: We wanted it to enable many small working dog groups to work together in a collaborative way, with data that were all uniformly coded so they could all have benefit of the estimated breeding values. That was its basic goal from the beginning. And that really hasn’t changed.

In the process, though, we’ve enabled IWDR to manage, to be a tool for managing, a complete organization, if they want to do it. It has modules now for enrolling new clients. For monitoring contacts with existing clients. It has another module for recording complete information from a veterinary clinic’s perspective for organizations that have their own in-house veterinarians. So it’s a really complete record keeping system. Now pretty well complete, at least for meeting the needs of guide and service dog organizations.

9:48

JH: That is actually more functionality than I knew that it had. And as you were talking, it suddenly dawned on me that I don’t think we ever said what IWDR stands for, and I’m pretty sure that’s going to be in the title of the podcast episode so people should be okay with that. But why don’t you go ahead and say what it stands for anyways?

(laughter)

EL: Well, we cast around for a while for an appropriate name for this database. But I still remember when I was out on an exercise walk arriving at the notion it was IWDR: the International Working Dog Registry. And at that point we were also considering IWDBA as the organization that should own IWDR. So, all the IWDs just fell together, and now people get totally confused about who owns what and what is IWDR, what is IWDBA? Are they the same? Are they different? Is one an organization and is one a database? So yes. So IWDBA, the International Working Dog Breeding Association, is the organization that owns IWDR, the International Working Dog Registry.

JH: So the important way to remember it is the one with DB in the name is not the database.

EL: I guess. (laughter) That’s true.

JH: Surely IWDR is going to take over the world and own everything at some point.

EL: Well, it still is the only database I can, I am aware of with a worldwide focus that is intended from the get-go to enable dog breeders to use estimated breeding values.

JH: Yeah. Which is incredibly valuable. And so theoretically, it can build EBVs for health traits, because you said you’re collecting information from veterinary records. And can it build EBVs for behavioral traits as well? How are you collecting behavior information?

EL: In the guide and service dog world there’s a tool known as the Behavior Checklist, which has been slowly adopted by a wide number of organizations as a tool for scoring aspects of behavior. And that’s the tool that we have primarily built into IWDR.

There’s a second tool that we don’t yet have programmed, but there are a few organizations using Dr. James Serpell’s Canine Behavior Assessment Research Questionnaire, also known as C-BARQ. And the genesis of both C-BARQ and the Behavior Checklist go back to the Seeing Eye in the 19-, well, no in the early 2000s. I think the first paper, Dr. Serpell, published on the C-BARQ was in 2002. But in order to publish that paper in 2002, of course, he had to do a lot of work beforehand. And a good bit of that work began with a project at The Seeing Eye where The Seeing Eye wished to gain information from puppy raisers. But they wished to gain that information in a coded format where it was standardized across all puppy raisers. So we could use those data for summary and to look at trends and so forth.

So The Seeing Eye actually engaged in a collaboration with Dr. Serpell to try and crack that nut, if you will. And so he did. He developed first a questionnaire that was a predecessor to the C-BARQ and decided that it worked so well he thought he could build one that was more general and would work for any canine/human interaction situation where a dog lived with a group of people who would come to know over time the the idiosyncrasies of that dog quite well and they would be able to describe that with answering these 101 questions that make up the C-BARQ. And so he and a collaborator of his, Yuying Hsu, took up the challenge of developing C-BARQ and that became the C-BARQ that is currently the validated tool for use by the general public. 

But part of the validation process for developing C-BARQ was to use a, what’s now a shortened version, or the original version, of the Behavior Checklist as a tool for the trainers of The Seeing Eye to provide independent scores on a different metric of dogs that came back to The Seeing Eye after having been scored by their puppy raisers using the C-BARQ. And his objective in that process was to see, did these trainers who clearly know dogs and dog behavior rather well (they’ve been senior trainers at The Seeing Eye for, you know, decades in some cases) did they see the same behaviors? And could they describe those in an independent way using the Behavior Checklist that were recorded for those same dogs using the C-BARQ. So those two tools kind of were developed hand in hand. But the C-BARQ has received the most attention because it was the validated tool and he had other methods of validating it that are recorded in his papers. But the C-BARQ was intended in the beginning just as that one component of the total validation process for C-BARQ.

The C-BARQ, though, at 101 questions is really quite long and rather laborious to fill out. And in general, puppy raisers are only being asked to use the C-BARQ, or to fill out a C-BARQ, when pups are six and 12 months of age. So they’re asked to do that twice in The Seeing Eye scheme. Other organizations use it in somewhat different ways, I think. But the BCL, the Behavior Checklist, is a much shorter list of questions. And at some point Jane Rusenburger saw the BCL as a potential alternative for scoring aspects of behavior in guide dogs. So she worked with Dr. Serpell to add some questions to the Behavior Checklist to cover components of guide dog behavior that were not covered in the original set of questions. And that is the genesis and the evolution of the Behavior Checklist that is now a part of IWDR.

And we now have enough records gathered on IWDR that we’re in the process of doing statistical analyses of those data, looking at all sorts of questions that can be answered, such as, can the 52 items that comprise the current BCL be reduced to a smaller subset of factors that describe aspects of behavior in fewer components than 52 questions? That’s one, one part of what’s the work that’s being done. Then we’re looking also at test/retest…. Well, what’s the…

JH: Reliability?

EL: Yes, reliability. Thank you. So if a person scores a dog today using the Behavior Checklist and they see that same dog three or four days later, or some few days later, and they score that dog again, do they score similarly? There’s been some work already done preliminarily to indicate that yes, the skilled observers do score the dog similarly. And of course, the dogs on day two, whenever that might be, are not the same, in exactly the same circumstance they were in on day one. So there will be variation, but you expect to see similar aspects of behavior if the tool is going to be useful. And so we’re attempting to do all those analyses and we’ve got help from skilled scientists elsewhere in the world helping us do those analyses.

19:48

JH: Yeah, I’ve been lucky enough to be on some of those calls and watching that progress. It’s really exciting. That’s gonna be a great paper when it comes out. I’m really looking forward to it. 

So people, by the way, who want to know more about the Behavior Checklist, or BCL, I did an interview with Jane Russenberger on one of my earlier podcast interviews and she talked about the BCL a lot. But so basically, in summary, we’ve got the C-BARQ, which is survey-based, already validated. So you’re using that where surveys are what is on offer. And then the BCL is something that is more intended for someone to be actually trained to use. And they both have their experience with the dog over however long they’ve known the dog and then also they often, or maybe always, have an in-person temperament or behavior assessment where they actually go do an in-person test with the dog generally? Is that right?

EL: Generally. That’s how Guiding Eyes uses the Behavior Checklist. And that’s how Jane has been teaching other guide dog schools to use it. But it’s a scoring tool. I think it’s important to recognize that it’s simply a method for describing aspects of behavior in a uniform scoring scheme. And there’s not necessarily any right or wrong score, necessarily, for each question. But if a dog exhibits a behavior that’s on the less desirable end of the scale for a particular, for that behavior, then that’s probably a dog you don’t want to continue working with for that particular job. But that dog could be well suited for some other job. And so the tool has that utility to it.

But yeah, it’s used in a variety of settings. Guiding Eyes actually uses it for scoring puppies in Puppy Test. And that’s a very defined protocol that has evolved at Guiding Eyes over at least a decade and a half, maybe two decades. Then they also use it when they observe dogs in what they call “walk and talk” situations. And Jane probably talked a lot about “walk and talks.” And then it’s used again when the dogs come back to begin training. At Guiding Eyes at least they all come back on the same day, and in a very well, predefined and well defined “in-for-training” test. The dogs are put through about a 15 minute protocol to see how they react. And it’s very standardized, and so it removes that source of variation to some extent, about variation from different test settings, although from month to month, there may be different people to some degree doing parts of the test. It’s a standardized protocol. So the Behavior Checklist has utility on all those situations. But it can be used by anyone who’s been trained to use it, observing a dog in any kind of daily activities. So, I think it has general utility.

JH: That is good to know. I had been thinking of it as more specific for having that in-person formal tests, but that’s good to know that it would be, it could be useful more generally. So then so, so then you have these behavior assessments for the dogs, which people can use to decide whether to go forward with them or not. But then because those are entered into the IWDR’s database, you can generate EBVs based on them to help schools make breeding decisions. So have you generated EBVs yet for health or behavior traits?

EL: We began calculating an EBV for the Pennhip score of hip quality about a year ago now. It’s been almost that long, I think. And that’s updated every night by routines that run in the database. We have an animal breeder working with us at the moment who is developing the workflow pipeline to calculate breeding values on other aspects of health and behavior.

For health, he’s working at the moment on an EBV for an extended-view hip score. And that’s an interesting one because worldwide, there are multiple schemes for scoring hips in the extended view, much like the OFA score here in the US. But there’s an FCI score. There’s a BVA score in the UK and Australia/New Zealand. All of those are different than the OFA score. But all of them intend to provide a clue about which hips have conformation that the radiologists believe is the most desirable and which hips are already showing signs of osteoarthritic change.

And our approach with those has been to give them a score code, which takes the best and the worst and those intermediate levels and re-codes them in a way that we think reflects the same general information. And so on a one to five scale we have hips from the worst at one to five that are best. And the preliminary analysis of that indicates that it’s something close to 30% heritable and certainly a trait that should respond to selection, provided that organizations are giving us all the data on most pups born in a litter. 

That’s a really key point that many organizations fail to fully understand. They often want to do radiographs on just a few dogs. If hip dysplasia is a problem they will radiograph the dogs that are their prospective breeders with the notion that if they choose the best ones from those, they will be choosing dogs that are less likely to transmit genes for hip dysplasia to their offspring. And that might be true, but it can be very misleading because a trait like the extended-view hip score, which is heritable in the 25 to 30% range is a trait that requires a lot of information about not just a single dog’s phenotype, but also the phenotypes of many of its close relatives.

If you look carefully at the OFA website about how they recommend breeders use the OFA hip score, they will say you should choose a dog with good or excellent hip score itself, that comes from a litter where most of the dogs in the litter have good or excellent hips, and comes from parents with good or excellent hips, and grandparents with good or excellent hips. If you put all that information together, what you really said is that’s a second way of trying to assess the value, or the quality of hips in a single dog’s relatives, by just kind of a “back of the envelope” calculation of an estimated breeding value.

But if you really look at the OFA database and you try and do that with the records that are published in the publicly accessible records, what’s missing is most information about dogs that are dysplastic, that don’t have good hip scores, because owners refused to allow that to be published. And you get a very biased reading of what the relatives of a dog really look like with respect to hip quality. So it’s really important to have recorded in the database, and IWDR in this case, information about as many relatives of a dog as possible.

So we continue to try and push that point and we talked to our subscribers about the need for doing that. And some of them are beginning to radiograph all the dogs born in every litter. I was really lucky at The Seeing Eye that that was a point they understood early on, and they regularly evaluate all dogs that come back for training. So, and they make that evaluation for every trait that’s an important health trait.

JH: Yeah, it’s important. The computer can’t figure out what’s going on unless it has all the data, right? You can’t just give it the good points. It has to understand the bad points, too. And that’s… I think a lot of people still feel like traits like hip dysplasia, they sort of are still thinking of it in simplistic terms, like how Mendel studied the peas. That there’s just a couple genes. But there’s a lot of genes interacting and to try to piece that out that computer really needs to see the good, the bad, the stuff in between, in order to get you the information.

EL: Exactly. And I, when I have the opportunity, I often make the statement that the most valuable information to have with respect to improving hip quality is which dogs are dysplastic. That’s often the piece of information that’s missing. If you look at the calculation of EBVs for hip quality based only on shades of normal, you get a quite different view of the presence of hip dysplasia in the population than if you could look at all the scores of all dogs, including those that are the worst dysplastic.

And one of the big criticisms of the OFA over decades now has been their lack of information about dogs that are dysplastic. Many dog owners will have the radiographs taken, then in the veterinarian’s office they will look at the radiograph. And if the dog is clearly dysplastic, they’ll choose not to even submit it to OFA. And that’s really doing a huge disservice to the breed as a whole. Because that’s the really most valuable piece of information. But it’s only valuable if it’s put into a scheme where that becomes available to everyone.

And a beauty of the EBVs is if all the data are recorded in a database, then the algorithms that calculate EBVs take those data points that are both good and bad hip quality data points into consideration in calculating an EBV. But when it’s all said and done, and the results are being reported, there’s nothing that says the dogs with poor EBVs have to be reported out to the public, per se. You just don’t offer those dogs for breeding. But individual dog owners know, “Well, this dog had a really good EBV for hip quality, so that’s one that I can promote as a good breeding candidate” and why. And if the EBV range is regularly reported, then everyone can see where that dog’s EBV lines up, along with all others. But they don’t have to know who are the bad dogs that made this dog look so good. They just have to know that this dog looks really, really good. And so they can use the information without having specific individual knowledge of, or specific knowledge of individual dogs that had poor hips.

JH: And I don’t want to stop talking about hip EBVs. But I just want to take a second to ask. So what you’re saying is that the IWDR provides the ability for people to upload information which is useful in calculations but which is not publicly displayed for others to see?

EL: Mmhmm. Right it does. And so the way IWDR is organized, the data are owned by the organizations who are the subscribers. And there’s limited information available on all dogs in the database to anyone who is a subscriber. That information is mostly about who is this dog, who owns it, who are its parents? And our primary reason for doing that is to provide information, or to provide subscribers with enough information, to avoid the entry of duplicates, duplicate records, when they start filling in the pedigrees of dogs that are currently available. And they wish to go backward and say, “Well, this Tom was the sire of this dog and so does Tom already have a record in the database?” And they go searching for Tom. We want them to have enough information available that they can find Tom. So that’s our reason for making all that information available on all dogs. But it’s very limited. Its birthdate and that kind of information. And it’s primarily to enable identifying ancestors.

JH: So that’s going to be an important thing for people who want to put their information into the IWDR to understand, that not everybody can see everything about their dog. That they can upload sensitive information, and if they want to keep it private, they can.

EL: Correct

JH: Wanted to make sure to call that out because I think that’s going to be important to some listeners.

Okay, so going back to the hip EBV. So you have pulled together different hip assessment scores from different parts of the world, and have at least enough information about good and bad, though it has been challenging to get the bad but enough information about the good and the bad to actually generate hip EBVs. And was this for just one particular breed of dog or multiple breeds?

EL: At the moment we only have adequate information to do this on Labrador Retrievers. We’re close to having enough information to do it on a small group of German shepherds. Those are the only two breeds that are so widely used in the guide and service dog world that it makes sense for doing all the work required to calculate EBVs. We could do it on any breed that wanted to bring all the records they could assemble on a breed into the database. And over time, that might evolve. But at the moment, we only have mostly records on Labradors and German shepherds. Some golden retrievers, and there are organizations beginning to use golden retrievers in larger numbers that will warrant calculating EBVs for some traits in goldens as well. That’s down the road.

JH: So how many dogs, or what kind of a depth of pedigree, do you need to be able to calculate a trait like that?

EL: Traditional pedigrees, if they go back three or four generations, are probably sufficient for tying together the information on relatives through ancestors so it makes a meaningful EBV. I’ve looked at some of the records in IWDR, and there are dogs in the database with pedigrees that go back to dogs that were alive in the 1930s. So we’re looking at like, 80, 90 years ago those dogs were alive, and those are now like 20 generations back or 30 generations back. And you don’t need to go that far. But this organization had records already in their old database that went back that far. And so I’m not going to throw out records just because they’re old, and…

JH: They had hip data on those dogs that you could use?

EL: No, they didn’t have hip data. But they’re just ancestral pedigrees. They do contribute to the accumulated inbreeding over generations. And all those records are taken into account by the routines we use for calculating levels of inbreeding so they do play a role.

JH: Well, so if a breed club were to come to you and say they had a particular problem, a disease or something, that they wanted to breed against, and they had some number of dogs where they had a depth of pedigree on the dogs and a depth… But they’d also have to know yes or no for that trait, right? Whether the dog had the disease or not? About how many dogs would you want them to hand to you with what sort of relatedness? Like if they were just to come to you and say, “We want to do this and we have the money to pay you to do this.” Can you give me an order of magnitude? Just because I feel like people listening to this are going to be thinking,”Well, my breed has this problem and I want to go to my breed club and talk about whether this can happen.” So what would you have to say to them?

39:34

EL: We would need records on, in the ballpark of, at least 1000 dogs to make it meaningful, and that would be 1000 related dogs. They don’t have to be closely related necessarily, but they would need to be related. If we’re going to use traditional pedigree relationships as the method for defining who’s related to whom and their degree of relationship. 

If they were funded well, and they wish to collect DNA and use snip marker arrays, then an alternative is to define relationships among dogs using snip marker arrays. And that technology also exists and is a capability of the software we’re using for calculating EBVs. But in that case, they would be purely GBVs: genomic breeding values. But it is possible to define the relationships based on DNA markers. And so if they don’t know who are the pedigrees, although most purebred breed clubs would know the pedigrees, at least to the extent that no dogs jumped the fence and sired puppies, in the dark of night, unknown to any humans. (laughter)

JH: That never happens.

EL: Yeah, right. So 1000 dogs that are related, with most of those dogs being phenotyped for the trait of concern, would be a number that we could start doing something with, I think. But it would be really important to have the phenotype recorded on most all members of every litter where an affected dog occurs. It doesn’t do any good to just know who are the affected dogs, we have to know who are the dogs that are also normal. 

JH: Yeah, for sure. So that, you know, in the case of relatively late-onset cancer that you wanted to breed away from, you would need to have a bunch of dogs in there that you had actually checked in and found out that the dogs had lived to 10 or 12, or whatever, and died of something else.

EL: Right. There is technology in the animal breeding world to fit models that are known as survival models. That’s where the age of an animal, or the variable being analyzed is “age of the dog at this point.” And if you know the dog is five years and three months old, you use that as its current… We call that the dependent variable. But it’s the variable being analyzed. And if the dog is still alive and known to be disease free, then a censoring variable is added into the equations, so that you know that the dog is still alive and healthy. But if the dog is now diseased, you can code it as being diseased. And if it died at some point for reasons other than this disease, then it’s a censored record. You just know that it was alive at that point, but it died for some other reason. And you don’t know what its status was.

And so that becomes a less useful but still a piece of information that can be used in the modeling to identify the dogs which are least likely to be affected by the disease and which, if it’s an age limited something like a cancer that generally is not seen in dogs younger than six years of age like some of them are, then then you’re looking for those that live to be 10 and 12 years of age and are cancer free. But you can’t afford to wait until all dogs are that age to do your analysis. You have to do the analysis using the data that exists today and as new information comes into the database when you update it next week, things will change based on the addition of new information that accumulated over the intervening week. And so those procedures are all well known and can be used. And, again, those can be done using either traditional pedigrees that define relationships, or they can be done using snip markers that define relationships.

Or a third alternative is a hybrid where some dogs are known by standard pedigrees to be related, while a smaller group of younger dogs usually are defined by snip marker relationships that have been done on the more recent dogs so you don’t necessarily have to have DNA on all the old dogs in order to use their relationship information. And this is all technology that’s now being widely used in the dairy industry, for example, worldwide actually. And the software we’re using for doing the EBV calculations is exactly the same software that’s being used in the dairy industry.

JH: Which is the dairy industry has made amazing progress on things like milk production using exactly these technologies. So they’re way ahead of us. 

EL: And they’ve got these huge databases that are very similar in scope to, or were when they were a young database, similar in scope to IWDR. But where IWDR now contains data on about 60,000 dogs, the dairy database for the US, the last time I was aware of the number which was fairly recently, was a database holding records on 60 million dairy cows. So the orders of magnitude are a bit different.

JH: That’s what happens when you have government support and government funding… It’s something we have to deal with in the dog world is that we just do it on our own.

EL: And that’s true, although the beef industry is similar, is in a similar situation, but they have had very little, if any, government support. So it is possible, where a group of animal owners decide it’s worth working together. And that’s really what IWDR is intended to promote is collaboration among people who have somewhat like-minded goals.

And I think it’s also important with respect to estimated breeding values to say that EBVs describe the germplasm that’s available. They don’t define what the goals of selection are. They simply describe the germplasm that’s available for selection. So a dog that’s right for an organization over here with a particular objective, does not necessarily become the right dog for an organization somewhere else in the world with a different objective. 

And I think people understand that, but they often think, “Well, I need the best. The dogs with the best EBVs.” Well, best is in, kind of like beauty, the eye of the beholder. And so you have to know what your goals are before having the breeding values to describe the germplasm you wish to select will really make any sense.

JH: Well, for sure. I mean, I think they’re what is best for one group of people is not best for another group of people. So in terms of behavior, one group of people might want to select super independent dogs that can go work without a whole lot of handler intervention. And another group of people might want to select for super biddable dogs who are very attentive to their handlers, right? And so those could be the same trait going and just selecting in opposite directions, but they could still collaborate to build the EBV to assess the same trait. They would just be going in different directions with…

EL: That’s correct. 

JH: So, all right. So we talked about how a breed club theoretically, or some group of purebred breeders, could come and say, “Well, we have a bunch of dogs with or without this trait we have to have, both with and without, and we have pedigrees for a bunch of them. And then we have maybe some dogs that are not as related to the other dogs, but we are willing to pay for doing a “snip chip,” so a genomic assessment on those dogs.” So they could have some sort of a mix of some closely related dogs by pedigree and some other dogs not as closely related, but assess their relationship with snip chip, and they could go and presumably get an EBV. So, which is fantastic. But then there’s also the group of people who don’t have dogs that are as closely related, right? So maybe they’re not managing purebreds. Maybe they’re managing mixed breeds or they’re outcrossing a lot. So EBVs would be less useful for them, right?

EL: Yes, probably. But EBVs are primarily a tool for capturing what’s transmitted from parents to offspring, in an additive genetic fashion. And additive is kind of like compound interest. It accumulates over generations of selection. If crossbreeding is producing the animal that is of primary importance to your breeding program, then what you’re capitalizing on in crossbreeding is restoring unique combinations of genes with every new cross, and you get the most diversity in restoring those genetic combinations by breeding two purebred parents of opposite breeds, or of different breeds, not opposite, but different.

So I know a common cross is the Labrador retriever with a golden retriever. And if you go back 120 years those two breeds were diverged in the UK from common sources. So there is some common germplasm in those two, but they look quite different today. And so, if you’re going to cross those two, you’re doing that to capitalize on the unique gene combinations that make goldens, goldens and Labradors, Labradors, but you’re wanting one of each in the offspring. And to maximize that you need to repeatedly breed purebred parents of different breeds to produce these F1 crossbred offspring.

If you’re going to move beyond the F1, and you want to use an F1 female, for example, as a parent, now you have to decide, “How am I going to choose a mate for her?” Are you going to go back to a Labrador male? Are you going to go back to a golden retriever male? Are you going to choose a male of a third breed? Or are you going to choose a male that also is an F1?

If you choose an F1 male to be the mate of an F1 female, now you’re producing what the geneticists would call F2 offspring. And now you’ve essentially blown apart the unique combinations that you created in the first cross. And now you’ve got a wide range of variation in the offspring that are going to produce some which are absolutely terrific, others that are in the middle with respect to whatever traits you think are important, and you’re going to be producing some that are way less desirable than anything you’ve ever wanted.

So in terms of crossbreeding, or producing mixed breed dogs as a production scheme, generally people are gonna settle on trying to find purebred parents of different breeds that produce the crosses, I think. But, some people may be focused on trying to produce a new breed by doing repeated crosses of various schemes that produce an animal with a phenotype that doesn’t currently exist. And they call that a new breed, like a Labradoodle, or something like that. And they can do that. But those will be happening in such small numbers that estimated breeding values won’t be of much help, because you just can’t get enough data on the large enough population of dogs. 

And I think this is an opportunity to make a point about the forces that change gene frequency. Gene frequency is what has to be changed if you’re going to produce genetic change over time. And there are four forces that can change gene frequency. There’s mutation and migration. And then a third force is chance. And those three, really humans don’t have much control over.

If mutations occur in the genome of a sufficient number of dogs, and if that mutation is advantageous in some way through natural selection, then it will survive. But if it’s not advantageous, or it occurs only occasionally, then it’s likely a mutation that just shows up and then it disappears. And you may not even realize it’s there. Or if it produces a really bad phenotype, you may understand that it’s there, but it doesn’t have a selective advantage or it’s not linked to something that has a selective advantage, then mutation isn’t much of a force to be reckoned with in gene frequency or changing gene frequency. 

Migration is simply the process of bringing in individuals to a population area that are completely unrelated and are from another part of the world that are of that same species. So the animals that are brought in have a different genetic makeup than the animals that are local. It’s somewhat a migration event when Labrador retrievers from Europe that are descendants of old lines that exist only in Europe, are brought to the US. That’s a migration event, and those dogs will likely come in with genetic makeup that’s different from the dogs that exist here in the US. So mutation and migration are those two forces.

Chance is simply the event that controls what happens in meiosis as gametes are formed. And it’s also a component when fertilization happens: Which sperm actually was successful in fertilizing an egg? So mutation, migration, and chance are the three forces that humans simply have very little control over. Although migration to some extent they do if they choose to bring animals in. 

So the big force that can produce genetic change when it’s used over time is selection, and selection will only be a force that can operate when you’re fortunate to have a large enough population of breeding animals that you can overcome the effect of chance. Chance, in a small population, will be the overriding force that produces change in gene frequency. And you don’t have any control over whether chance will produce animals you desire in the good direction, if you will, or animals that you deplore, that are in the bad direction. And so that’s the primary reason you have to begin working together: So you’re working in a large enough population that you can overcome the effect of chance as a force changing gene frequency. 

That’s kind of a related piece of information to your question, but I think that’s an important point that I hope I’ve adequately explained in a way that most breeders can understand.

57:57

JH: Yeah, for sure. So big populations make chance less powerful. And so genetic drift is the other thing I think that we would, and the other term we would use to describe that.

EL: Correct. Chance and genetic drift are the same thing.

JH: Yep. It’s just if you have a small population and something happens by chance where you happen to lose some good versions of genes, such a small population, you don’t have them anywhere else to replace. But if it’s a big population those small chance, you know, losses are buffered by having stuff elsewhere. 

EL: That’s correct.

JH: Yeah. So it sounds like at least with F1s, if there were a large enough population of F1s, then it might be that EBVs could be built up. But there are fewer of those first generation crosses out there. 

EL: [indistinct]… true. 

JH: But something that could be interesting to look at for the future.

EL: There are techniques for calculating breeding values in cross bred populations that do take into account dominance, but dominance, genetic dominance, has to be reconstituted with the creation of every new generation. It will be reconstituted, and so it’s not like additive genetic effects, which are regularly passed from parents to offspring. And so that’s the reason most EBVs are done for purebred populations. You’re capitalizing on accumulating genetic improvement generation over generation. So.

JH: But if people who were crossbreeding dogs were interested in using the IWDR, there’s other benefits to it for them, even if EBVs were a ways off for them. 

EL: Sure

JH: You talked about some other ways in which you can actually help you manage your breeding program.

EL: Well. Even if you are working with a breed without sufficient numbers for you really using EBVs, at least IWDR offers the ability to record data from any place in the world using a smartphone, or a tablet, or a laptop, or desktop. Anything with a web browser. And it will record those data and capture them in a location in the cloud, the internet cloud, that ensures that once you’ve pressed “Enter” and the signals have gone from your device to the server, the server has it. So we’ve not yet experienced a time when that failed to work. Sure, glitches can occur and things could happen. But generally the servers have been up when

A few weeks ago we had a storm here in my part of the world that took out power for us for five days and the internet for seven. So we locally were unable to do anything. But others elsewhere in the world did not know that we were without power and had no internet connection. And the servers just kept running and they did their job and nothing was disrupted with IWDR because they live in a huge data center that’s powered by backup generators and is a protected location. It’s not even based in the US. And so the servers just kept doing their job and IWDR kept functioning.

JH: So reliability. Very important.

EL: Yeah, reliability and also the presence of backups. We make backups every day and at the moment we’ve not yet exceeded our ability to store all the backups we’ve ever created. So we’ve actually got backups that go back for a number of years.

But the other advantage of using IWDR is that the data will be uniformly coded. And so when you decide it’s desirable to do some summaries of those data, whether it’s with EBVs or just looking at the incidence of phenotypes and so forth, then the data are uniformly coded and were captured from the beginning where summary analyses were a part of the the objective for storing the data they began with.

And I can’t tell you how disappointing it is to find organizations who built these recordkeeping systems where most of the data are captured as textual comments. And you can’t believe how many different ways people can write, “Labrador retriever exhibited some undesirable behavior, was afraid of tricycles or something.” If you wanted to know how many dogs did we have that were Labrador retrievers afraid of tricycles, you would have to search through text recorded in a myriad of ways in order to find all the dogs that are Labrador retrievers, afraid of tricycles.

Well, that’s not going to be a problem in IWDR because from the beginning IWDR captures most of its information in coded lists. And so the presence of an attribute is noted in the records by a numeric code. It’s not noted by the presence of free text that someone has typed into a comment box.

1:04:01

JH: And it provides… And so you have done all the thinking ahead of time of what all of those various pieces of information that someone might want to save about their breeding program is… right? All the different health traits; all the different behavioral traits. As we said, there’s behavioral assessments in there. Just all those different pieces of information. You don’t have to set up your own plan for what information to save. It’s just there. 

EL: Yes, although I would modify your statement just slightly by saying we probably don’t have all the important health conditions for all dogs of the world. We have all the important ones for dogs that we are aware of, and primarily the Labrador, German shepherd and golden retriever populations. But if a different breed were to come into IWDR with things that don’t currently exist in our coding schemes, of course, we would add those. So we could accommodate them. But they would be added by us into the coding scheme so that the text which describes the code would be uniformly recognized by all members of that breed club or group who wish to cooperate. And so the dogs would be uniformly coded as those records were being created for a particular health condition, whatever that might be.

JH: Um, so are you interested in having new breeds coming into the IWDR? At this point?

EL: Yes, we’re interested, very much interested because IWDR, is intended to be a database that serves the canine breeding world.

JH: You do have “working” in the name of it, but it’s not actually really limited to working dogs, right?

EL: That’s true. It’s not really limited by anything inherent in the database itself. But if we were to bring in a breed that’s not considered a working breed, then there would be some additional work we would need to do to properly accommodate that breed, and whatever are the unique things they record, particularly about aspects of behavior. And in those cases, or in that case, the behavior checklist probably would not meet their need. They probably already have something developed for that breed that we might need to incorporate into our data entry scheme. 

And I should also point out that if a breed already has an existing database that they want to continue using, we could help them in the calculation of EBVs if we work together to set up a mechanism for adding new information to the IWDR using what’s known as an API call. We are in the process of building an API, or defining an API, to access records in the database for validated subscribers. And so new data could come into IWDR through an API call, which would not require making a major change to a group of users who are already accustomed to using an existing database. But those data would come in then in a way that would code them in the IWDR structure. So the records captured in IWDR would still be uniformly coded.

JH: So we could think of an API call as a bridge so that you have two databases. And there’s the information that a breed has been saving in their breed database for however many generations, and separately there’s IWDR’s database and we can think of the API call as a bridge between the two so the two databases can talk to each other. Whereas the other alternative would be to sort of shut one database down and suck all its information into the IWDR. And that may not be what everybody wants, which is what you’re saying.

EL: Correct. Exactly.

JH: So we’ve talked about groups coming in and bringing large numbers of dogs in and I guess my perception of what you’ve been saying is that if an individual breeder has an interest in the IWDR, they would need to come in through a group and not not on their own. Am I right about that?

EL: Yes, we’ve really built IWDR to accommodate a group of breeders, as an organization. And if we were to change that scheme, then we would have more work to do that we simply don’t have time right now to try and accommodate. So… 

JH: With your three people.

EL: Yeah, with our three people.

But if we had a group like that, then we’d certainly love to talk to them and begin to get a feel for what we could do for the group. But a single breeder that’s of a breed that we don’t currently already have in IWDR we probably are not the right place for them to begin putting records into a database. There probably are other places.

I know there’s an online system out of Australia that at one point I saw had been describing their database as having over a million records in the canine database. But it was a database for organizing and displaying pedigrees. I think they probably did inbreeding calculations, but I don’t think they were doing the work that we’re doing to store health and behavior data in a uniformly coded scheme.

JH: Well, the Functional Dog Collaborative is all about collaborating and pulling people together into groups. So I’ll leave that there and say maybe we could all talk more about that on the Facebook group. 

Well, is there anything else about the IWDR that you wanted to say, Eldin, that I have failed to ask you about?

EL: Not that I can think of. We’ve been talking for over an hour. (laughter). I think you’ve been extremely thorough. I may think of something tomorrow that I didn’t say, but if I don’t think of it until tomorrow, it’s much less important than what we’re discussing today.

JH: So if people wanted to just go… Is there a website they can go to to check out the IWDR and learn more? They’re not able to really see the full thing without being a subscriber, but

where could they go to…

EL: So if you go to www.iwdr.org, you will be at the website where IWDR is described. If you click on the links that are in the menu options that are in the upper right area of that homepage screen, you will find a link for the IWDR manual. If you go to the IWDR manual (and I think it’s chapter three) there’s a set of short videos that show you how to do various things in IWDR. If you watch some of those videos you will get an immediate feel for what the data entry screens look like. And you can listen to Jane’s description of how to accomplish a particular workflow task, whatever that might be, which is the focus of that video. I think the longest of her videos are 20 minutes, but most of them are five to seven minutes, so they are short videos describing how to do various things.

Then there’s also a collection of chapters describing in text and with screenshots how to use the IWDR, so you can gain a lot of understanding from that, and that would be a good place to begin getting a feel for what the database looks like and how it functions.

JH: Fabulous. And again, if there are groups who are interested in learning more about this, I think Eldin would encourage you to reach out to him. But I also would encourage you if you are not quite ready for that step to reach out to me, and I’m also happy to help talk to groups about what’s right for them.

EL: That would be great. And I really appreciate your being an ambassador for IWDR and telling others about us. I hope we can reach a point where many groups find IWDR to be useful, and we can create a community of users that help improve the quality of dog breeding by producing much healthier dogs over generations of selection.

JH: Yeah, that’s the goal and you have an incredible resource to help breeders do that. So thank you for making that be your passion project and putting this together in what should have been a blissful, relaxing retirement. Instead this is what you…

EL: I think retirement is a folly. I decided retirement is a good way to prepare to die. I’m not quite ready for that yet, so I’m… (laughter)

JH:…good for you…

EL:… I’m having a good time working on this and I appreciate so much the friendships that we’ve formed over many decades now working in the guide and service dog world and I cherish the interaction with people who are of like minded… Who have like minded goals and so my hat is off to all of them who pursue the production of better puppies in successive generations as they go about doing the things they do to help produce dogs that help people. 

JH: It is an amazing community.

EL: Yes.

JH: All right. Well, thank you again, Eldin. This has been fantastic.

EL: Thank you. Take care.

Thanks so much for listening. The Functional Breeding Podcast is a product of the Functional Dog Collaborative and was produced by Sarah Espinosa Socal. Come join us at the Functional Breeding Facebook group to talk about this episode or about responsible breeding practices in general. To learn more about the Functional Dog Collaborative, check out www.functionalbreeding.org. Enjoy your dogs.

Copyright © 2021 Functional Dog Collaborative. All rights reserved.