 Well, I promise to talk really fast, no seriously, the bedside. And do we have a pointer here? Yeah. Is this a pointer? So in the upper left here, I'll be talking about work done in our group at the University of Oregon. And the rest of the people listed on here are clinical groups that we have collaborated with and funding is listed as well. You'll note that essentially all these clinical groups are from Europe, one group from Canada. So we're looking forward to U.S. collaborations, and that's part of the issue that I want to bring up. So as I understand it, my task here today is to provide a perspective from the bench. I'd like to do that by describing two case studies that we have, oh, thanks very much, Carol. Two case studies of successful examples of collaborations with clinical groups. And then I'd like to discuss some of these gaps that Carol alluded to based on our experience working with clinicians. And then at the end, I'd like to touch on the Undiagnosed Disease Network as an example of a way that I think we can have very successful collaborations between clinicians and basic researchers. So the first case study is regarding Usher syndrome and gene discovery. So Usher syndrome is the leading cause of genetically-based combined deafness and blindness. The deafness is congenital, so present at birth it's due to sensory neural hearing loss and the vision loss is due to retinitis pigmentosa. So it's a relatively late onset progressive degeneration of the retina. This is a multi-factorial disease. We know of 11 different genes in which if there is a mutation, this causes Usher syndrome. It's a classic Mendelian recessive disease. Importantly, however, there are still many families that are unlinked to any known locus. And gene discovery is extremely important for genetic counseling in Usher syndrome. If you have a child who fails a hearing test at birth, you'll get some counseling. You could raise that child in the deaf community, which is a reasonable choice, but you may also want to consider cochlear implants. And obviously if this child is later going to be blind, sign language will be insufficient form of communication. And so having this knowledge up front can really help inform the decision by the family about whether to move forward with the cochlear implant and obviously would elevate the child on the priority list for receiving a cochlear implant. However, because the vision loss is very progressive, late and progressive, we don't have good means of testing vision in newborns. And so really the only way to know that this child is later going to have Usher syndrome is through genetic testing. And if you don't know what the genes are, it's pretty hard to test for them. So gene discovery is still extremely important. So we have developed zebrafish models of all the known Usher genes and are studying their functions. And I had a postdoc who was presenting a poster at an international conference. And a clinician, Hano Bolz, came up to our poster and said, I've got a bucket full of patients and families, and we're trying to discover what the genes are. Do you want to collaborate? So this is really sort of a one-off, happenstance interaction that we had. And this led to gene discovery. So they, that group, had done a whole exome sequencing of these families and came up with a candidate, the PDZD7 gene, which was a completely unknown function at the time. And they had picked up variants of unknown pathogenicity in several different families. So how do you move forward from this? How do you try to show causality that mutations in this gene can cause the disease? So we cloned the ortholog and zebrafish. We showed that the protein co-localizes in the retina and in the inner ear with other known Usher syndrome proteins. We showed that in the mutant loss of function of the, or in this case of morpholino, loss of function of this protein results in defective stereocilia and deafness in the zebrafish. And we also showed that it can lead to retinal degeneration. So interestingly, of the families that we've identified, in every case, the PDZD7 mutation appears in a heterozygous form with other known Usher gene mutations, either in heterozygous or homozygous form. So here we see a family in which the son is heterozygous for a known Usher gene, GPR 98, which normally would have no symptoms at all because this is a recessive Mendelian disease, but this patient presents with Usher syndrome. This family is even more interesting. They have two daughters, each of which is homozygous for a known Usher 2 mutation. One of the daughters carries an additional variant allele of PDZD7, and she presents with much more severe symptoms than her sibling. So these types of data suggested to us that PDZD7 may actually be acting as a genetic modifier. So to analyze that, we could use the zebrafish to model these genotypes using morpholinos. So in this experiment here, we use a half-dose, in other words, knocked down about 50 percent of the protein activity of the GPR protein and the PDZD7 protein. And of course, there's no phenotype. It's heterozygous. But when we combine these two half-doses, we see retinal degeneration as in the patient. And this would be the genotype of the other family where we have a homozygous, a full knocked down of Usher 2a, a half-dose of the PDZD7, and this exacerbates the phenotype, as we saw in the patient family. So this suggests, again, a genetic interaction. So we were able to show with pull-down assays that these two Usher, well-known Usher proteins bind to the PDZ domains of PDZD7. And currently, the model is that these proteins actually form a quaternary complex. And this elevates PDZD7 essentially to the status of being a disease-causing gene. And unlike Howard's example, shortly after this work was published, PDZD7 was added to the panel of genes that are tested in children who are potentially, who have Usher syndrome in which genes, they're trying to identify what the actual variants are. So this is a really successful example, I think, of how you can go from the bench, from the bedside to the bench and back again. But there are still some gaps. I think that this successful example illustrates. First of all, where are the missing homozygous and compound heterozygous patients? Is this simply the fact that it's embryonic lethal? The model organism data would suggest not. We've subsequently made mouse mutants. They also are perfectly healthy, happy animals. So it seems to us more likely that there are probably, that the patient pool is just too small. So there are patients out there that we haven't found yet that are homozygous or compound heterozygous for mutations of PDZD7. So why haven't we found them? Well, is it because there's limited access to patient data? Is it perhaps that the clinicians are not telling us about the patients that they have? They're not testing them? Or worse yet, are they actually acting as silos and hanging onto their patients and not sharing? So let's move to the second case. This is the opposite type of experiment in which negative results can actually reveal incorrect diagnoses. And this example is Jubeir syndrome. So here we had a family, a consanguous family, that have three children with deafness. Due to the consanginity, we thought we could do mapping for homozygosity by descent. However, this was uninformative. We came up with no good candidates. So again, we relied on whole exome sequencing, again looking for homozygous snips because of the structure of this family. And we found homozygous mutations in AHI1 gene, which is a very well-known gene that's responsible for Jubeir syndrome. Jubeir syndrome, as you may know, is due to under development of the cerebellum and brainstem. It involves patients present with seizures, retinitis pigmentosa, developmental abnormalities, kidney and liver abnormalities. And these patients were deaf but had none of these other signs. Most importantly, the MRI was normal, and this is the key diagnostic for Jubeir syndrome, the abnormality of the midbrain in which it appears sort of like a molar tooth. So what's going on here? So we look at the mutation, the specific allele is a truncation of the protein right at the beginning of the very important protein-protein interaction domain. So most of us, based on our biochemical training, would argue that this is going to be a loss of function because this protein cannot bind to its binding partners. The mutations in the various databases that are known to be disease-causing cluster in more interminable region of the protein. So again, we made models of this, truncating the protein, introducing mutations in this region that's known to cause disease. And this produced a very strong Jubeir-like syndrome in the zebrafish, which is extreme ciliopathy, very consistent with other known Jubeir gene mutations. However, when we modeled the patient by truncating the protein, eliminating this protein-protein interaction domain, the animals developed relatively normally and did not show any of the traditional Jubeir syndrome phenotype. So this is really kind of a scary result, in a way, because patients that have this mutation would be predicted to get the disease when, in fact, they do not. So I think this has pretty important implications for the interpretation of exome data, particularly in terms of diagnoses, making decisions about treatments, and I think especially for the interpretation of variants of unknown pathogenicity. So we went back to the family on the basis of this, and were able to obtain DNA samples from other SIBS and found that, in fact, one of the other SIBS was homozygous for the same alleles and yet did not have the deafness. So I think together this really points out some of the strengths of even getting negative results. So together, these two studies, I think, illustrate several points, several gaps between the exchange of information, the exchange of information between clinical and basic researcher groups. Most importantly, I think there are very significant barriers for basic researchers to access patient data. So some of these barriers might be sociological. I think Howard sort of alluded to this. It's the different attitudes that clinicians and basic researchers have. Also there can be certainly limited access, and this has been our experience, to clinical records. We have found that it's extremely important to get as much information as possible about the patient phenotypes, and in these examples that I talked about there, it's been sort of haphazard interactions that we've had. So we've typically gotten sort of de-identified clinical records. But really, if we could obtain IRB-approved access to the full clinical records, we could look at the full suite of symptoms that the patients have and thus produce, I think, much better models. And I've touched on the idea, too, that there may just be very limited access to patient data. And I think that's a problem that's difficult, one that we should face and perhaps come up with some solutions. So in contrast, in the last couple of minutes, I'd like to just discuss our interactions with the Undiagnosed Disease Network as an example of a very successful way to move forward. But I think, as most of you probably know, the UDN is composed of seven clinical groups distributed across the U.S. There are two sequencing centers, Hudson Alpha is here at this meeting. There's a metabolomics core, and of course, most importantly, there's a model organism screening center, which flies in Baylor and zebrafish at the University of Oregon. And this group, because it is a network and a collaboration, means that we, as the basic researchers, have full IRB-approved access to all the clinical records. This network is identifying thousands of patients with undiagnosed diseases. The sequencing centers identify the variants of unknown pathogenicity, and then we generate models of those exact variants using then the wild-type human CDNAs to rescue the mutant phenotypes as a proof of principle for validating these variants of unknown pathogenicity as disease-causing. And this is really, I think, an excellent example of how one can have a free and open exchange of information between clinical groups and basic researchers. But obviously, it's limited just to two basic research labs. So I would say this type of model should really be dramatically expanded and make it available to basic researchers throughout the country. Before the network came along, as I said, our interactions were really extremely haphazard and sort of one-off, and this is really pretty frustrating when one has the best model out there, obviously, and wants to be able to have clinical collaborators to help identify the genetic basis of undiagnosed diseases. So in summary, I've given a couple of case studies here where positive results can help validate genes, where negative results can reveal potentially incorrect diagnoses. And I've brought up some of the issues about gaps, at least from the perspective of the basic researcher, and discussed briefly the Undiagnosed Disease Network. So I'd be happy to take questions. Thank you. Questions. Thanks for that great presentation. I wanted to just ask the question, the context of the negative results. Given the concept of incomplete penetrance, how do you weigh your negative results in the context of what you see within a pedigree, between pedigree, and this concept that some variants may have incomplete penetrance, depending on the genomic context, how do you deal with that issue? Right. Yeah. So that's a really good question. And I think in terms of the animal models, it's difficult. So the examples that I talked about, we're looking at very strong loss of function mutations, and essentially Mendelian inheritance. And obviously, a lot of the variants that are coming, particularly the ones through the Undiagnosed Disease Network, have other modes of inheritance. So I think you have to really deal with that on a case-by-case basis, and have as much information about the human genetics as possible, and then try to develop a model that will be appropriate. There are obviously going to be limitations inherent in any model system. My question is related to the same topic, and it's, for me, I believe in the case of the Jovert syndrome, I probably believe what you presented, and that case in which, despite being homozygous, being expressed is a great opportunity, from my perspective, to look at what environmental factors, for example, dietary fat, from my perspective, could be influencing the expression of that phenotype. I remember something related, like retinitis pigmentosa, in which apoE and omega3 could be playing a complex role there in the expression, so that's a great opportunity to learn more rather than to dismiss that as being a real finding. Right, right, very good point, and I think that's the strength and probably the insight of the undiagnosed disease network to include a metabolomics core, so it really makes sense to run not only the human tissues through the metabolomics core, but also to use the metabolomic cores to analyze the model organisms. So yes, as holistic approach as possible is extremely important, particularly in the case of these negative results. I think it's important to consider or think about the question of causality and penetrance as independent, although when you get decreased penetrance, it makes the study of the causality more difficult, obviously, but in the clinic, one could say, we know that this causes this disorder, we don't know how often it's penetrant, so the information may still be useful to the patient, even though we don't know the complete story on the penetrance. Good point, thank you. Just a quick comment that I think we can't expect that the medical record holds all the phenotype information, and I think we have to remember there's a patient in the bed or a person in the chair, and very often the information we need to correlate with the symptoms that the patient is having, only the patient knows, and so I don't know how we capture that data, but I think it's a very important issue as we go from bench to bedside and back. Yes, that is a good point. I think the point I was making is that whatever patient information is available, that should be made available to the basic researchers, too, because it's extremely frustrating to follow a path when you have only half the information. Dan? I can't see. I hear a voice. The one next. Yeah, so this last discussion sort of moves us a little bit from sort of straight Mendelian disease. When you start thinking about modifiers, that sort of moves us along closer and closer to more common diseases. I think you've already made a compelling case. The UDN might be a great model for how do we think about expanding the basic science clinical interactions for Mendelian disease, but how far down the line do you think we can go with this in terms of complex disease? Right. So it probably depends a lot upon the model. So you can imagine making double and triple mutants in mice, but it gets painfully expensive and difficult as you move along. And so perhaps other models, zebrafish and maybe other, yet other models would have some benefits there when we get to multifactorial diseases. Mark, a quick comment on aquatic cancer. And this is just in response to Deborah's comment about the patient. I think we do have models that are emerging through a Genome Connect and others that are looking at innovative ways to engage patients and families in this type of work. And I think that that's something that also needs to be considered in the context of this discussion. I think we'll stop there. Thank you. And move on to the clinical perspective on need for integration.