Polygenic Risk Score Prediction of Complex Diseases in Companion Animals

And those projects take a while , so it wasn't until 2017 that the first paper from my lab was published on that topic , but already , even in that very first paper , we'd started to do some analysis about the genetic contribution to the disease and how that could potentially be used to predict cases from controls cases from controls .

And so here we are today , some years later , with a much more sort of sophisticated understanding of the topic in general about genetic risk prediction for common , complex or polygenic diseases .

But also we can right now write a paper to bring more attention to hologenic risk score , how this quantity can be used in practical for the risk stratification in companion animal , like dogs , dogs .

So one thing is that I told Peter we need to discuss about the differences between breeds in terms of the RISC scores and how we can optimize our models , how we can develop a model . So I can say this article brings several different things to readers .

One thing is same as human production animal polygenic risk score can be used in companion animal as well for risk stratification . But we have a huge diversity in dog population , for example , in pets . So we need to consider this .

We need to develop our model accurately predict polygenic risk score as just dog's population , for example , or other pet population , so this quantity can be used for personalized veterinary care . This is what brings these papers to readers .

What sparked your interest in polygenic risk scores and why are companion animals so far behind humans and production animals ?

So already , like in our lab , as many other um uh investigators are working , is there , essentially , labs are um building up that to some degree , their own biobank , and I think one of the challenges for the future will be how to figure out ways of sharing biobank data between labs or institutions , because because , for sure , in this genomics general

field , what's possible or questions that you can answer are definitely related to the magnitude of the data set that you have access to .

My interest in polygenic risk scores really originated , as I mentioned , in looking as a clinician , looking at animals with orthopedic problems and particularly dogs with crucial ligament rupture , where , pretty quickly , any um clinician who works with some breeders or works with a lot of um uh trainers or field trial dogs and that type of thing , you come to realize

pretty quickly that there's more going on with this condition than just accidental injury . And so , as a sort of clinician , clinician , scientist , then obviously you're starting to ask questions well , what is actually really going on with this very common problem ? And so that's really what drove our interest in this topic area .

And I think that one of the challenges has been to figure out sources of funding that can support impactful work on new sort of big problems .

And I think the other other thing is still that that some of the work is still in an early phase , so it's definitely an area where production animal science is ahead of humans and humans and production animal science is ahead of companion animals , and in that sense I include like horses , as in companion animals , as well as like dogs and cats .

But it is starting to change and move forward , moving forward a fair bit , and I think the promise for the future is pretty bright a fair bit and I think the promise for the future is pretty bright .

And back to crucioligament rupture we know clinically and have known for a long time that this is a condition that's common in multiple different breeds of dog , and our academic papers have published on a small number of breeds , but principally the Labrador Retriever , because it's the most common until recently breed in the US which is commonly affected with this

condition . And so that was the sort of reason for focusing on the Labrador Retriever in the beginning . But we know from ongoing work that we're continuing to pursue that it is quite

challenging to do predictions across different breeds of domesticated animal or essentially populations of different ancestry , and we think that that can be overcome with some more research and more funding .

But it is a challenge and essentially it boils down to this point about genetic heterogeneity that although cruciate ligament rupture , as a prototypical example , is quite heritable , it's quite common in different breeds of dogs .

There's heterogeneity in the genetic contribution in the different breeds and so a data set that works well for prediction in one breed will not necessarily work well for prediction in another breed , and so solutions to that sort of scientific challenge is still needed .

So one thing can improve our prediction accuracy is a well-optimized reference population . We're always trying to have a transferability of our polygenic risk score across different breeds , so we need to have a reference population composed of many different breeds .

When we estimate SNP effects , this SNP effect could be representative of all breeds , representative of all breeds . So one thing could be improving a good reference population , establishing a reference population . Another thing is that with advances in genomic technology , we can have many layers of omics , information , genomics , transcriptomics and epigenomics .

Another thing is that how we can use these multiple layers of information , genomic information , to fit to our model and improve the accuracy . So I think we need to think about this area as well and provide

more information , more input for our models to improve the prediction accuracies .

So AI can play a role for providing us the most accurate input or phenotypes . On the other side , we have exploratory variables , so they have patterns . Ai models have shown they have a high ability to recognize the pattern of data , for example , nonlinear patterns .

So I believe AI can be used for detecting this pattern , to predict or empower our model for prediction .

So I think in the current era of science moving forward , good teamwork , I think , and good interdisciplinary collaborations are going to continue to be very pivotal .

And he looked at me and he said , lisa , I model the earth . And I was like , oh right , it's really respectful for different and it just excites you so it's easier to sit down and really get through a manuscript . But well done , I mean , you're an amazing human being and individual .

And I think I'm certainly true at UW-Madison , but I suspect that it's generally true in many other veterinary schools . And I think I'm certainly like projecting into the future development of thinking over what a curriculum should be , an ideal curriculum should be for veterinary students . I think that's an area that definitely needs some more reflection of .

Veterinarians are going to have bigger

roles and bigger exposure to sort of large data sets , and computer science and bioinformatics are going to continue to play a very important role in like veterinary medicine just in general .

I will hope your dean is not listening or you'll end up on the curriculum redesign committee listening , you'll end up on the Curriculum Redesign Committee , so that's certainly been a theme for us in terms of you know how we think about things here .

And so back to the work that we did with crucial ligament rupture in dogs . All of that initial work was done using the binary trait and each dog was either a case or control , and obviously , as a clinician , the next Labrador that walks into your office , if you flip a coin and call it as a case or control , you're going to be right half the time .

So heritability essentially highlights the potential for a predictive genetic test . So , for example , in the scenario where the heritability of crucioligamen rupture in the Labrador retriever was estimated as 40% or 0.4 , then with an ideal setup , genetic risk testing should be 90% accurate , because 0.5 and 0.4 is 0.9 .

And so I think the take-home message for veterinarians is that there's a strong linkage between genetic risk prediction and heritability , and so often the place that this work starts for a new disease or condition is to estimate the heritability . Anything maybe can add to that a bit more .

Our dog is at high risk , medium risk or low risk , so before any clinical signs emerge , before any clinical signs emerge , so we can use this value to manage to adjust , to change the lifestyle for our pets .

I remember when we published our genetic test for the first time and we announced through our Facebook page somebody wrote I am skating to get this test for my dog . I want , I am going to say polygenic risk score values for any disease for your pet is not scary , it's just caring . It helps you to help your dog .

It navigates you through the changing or adjusting your pet in terms of the daily activity .

And so even today , we're still regularly working in the clinic with owners who have a dog with cruciate ligament rupture , where they still have the perspective that the dog was playing ball in the garden a few days ago and became lame and developed the condition and so in their mind it's an injury situation , injury situation on , whereas the reality is , um , there

might have been some uh activity associated with with the rupture event , but the underlying um reason the problem arose is as because of intrinsic like genetic disease .

Um , in , obviously , just again , using that as an example , we have this paradoxical scenario where greyhounds are the fastest , most athletic dog you could possibly come across and as a breed , they're heavily protected against the risk of crucial ligament rupture .

So I think that's still the thing that owners should be aware of in terms of orthopedic , common orthopedic diseases in general , but cruciate ligament rupture in particular .

Really fascinating work and Peter , for all you've contributed to especially AJVR , but JAVMA as well and really supporting our journals . We really appreciate it . As we wind down , we like to ask a little more of a fun question .

So , Peter , for you and if you have it , you can show us what is the oldest or most interesting item on your desk or in your desk drawer .