 OK, so I was told that maybe a lapel mic would work better. Can you hear me OK? Good, that means I can wander. What was that? OK, I'll do it this way. Since I only have 10 minutes, I'm going to jump right in. There are different levels of analysis we're doing in a clinical laboratory from targeted exons that we may seek for known specific mutations. Many times there's a single gene that causes a disease that we could look at it. But we have now entered the period where we do gene panels. And some of these can get quite large. And so we're dealing with multiple genes, multiple variants to be able to work through. And we are very close to the time of a clinical whole exome type test. Traditionally, our regions that we interrogate are the exons, intron-exon boundaries, known intronic mutations, and if known gene regulatory regions. We do not do the deep intronic mutations unless something more is known. So we don't want it to come up with tons of variants of uncertain significance that we know we can't handle in the first place. So I wanted to bring out American College of Medical Genetics recommendations for laboratories that says that the laboratory must provide the interpretive information and a best estimate of clinical significance for the variants. So it is a responsibility of the laboratory. And so it starts in the clinical lab when we write the report for the patient that we try and give the best estimate and the best classification that we can. On the side here, this algorithm is actually with that paper. And you can see it's fairly complex. So I'm going to try and simplify it a little bit. Basically, we look to see if it's been previously reported. And if so, does it tell us whether it's pathogenic or benign or somewhere in between? And what is enough evidence to collect to be able to classify it as pathogenic? We need to realize, and I can't emphasize enough, that in a clinical lab, we're dealing with a lot of different diseases, a lot of different genes. Each of them have their own little quirks about it, I'd say. And we need to realize that if it's autosomal recessive versus X-linked, we're going to approach it differently. We're going to approach dominant versus recessive diseases differently. A simple, basically, if you have a dominant disease, you look for one mutation that is causative. If you have a recessive, you have to find the two mutations and show that they're on different chromosomes. If it has been previously unreported and we can't find it anywhere, then there are many times that we can say, this is expected to be pathogenic. It's a stop codon and nonsense mutation in the middle of the gene. And we will report it as pathogenic. If there's less certainty, we'll say it's suspected pathogenic, uncertain or suspected benign. And then we are getting to the area that we're doing some further classification on severity. Not often, but at times that we can talk about severe, moderate, mild, or even very mild mutations. An example of this would be pancreatitis, that we're looking in the cystic fibrosis gene CFTR gene to find mutations that do not cause cystic fibrosis but cause pancreatitis. And so we need to put together a severe mutation with a mild or a very mild mutation to be able to give a report with that. So there are times that we need to classify more than just pathogenic. We need to talk about severity. So this is our process right now. We check our internal databases. Have we seen it before? What have we classified it before? What evidence did we find before? So we don't have to repeat that work. Now, the problem here is that there's going to be differences between different laboratories because I'm going to see different variation than another laboratory. And so you may get differences between the different laboratories and we need to become more standard. I do look at local specific databases. And I'm including the biobasin with this. But again, there's differences between databases. Some of them are great. Some of them I really like. And other ones are only marginally useful. Some of the differences are that the evidence is given, that they may not give much evidence. We don't know how often or sometimes they're not updated as regularly as what we would need. And as Les pointed out, we do need to check original sources. Simply finding a mutation reported in a paper, one time I pulled it up. And that mutation was actually found in the control population, not in the disease population. But it was labeled as pathogenic in the database. We look at DBSNP and mainly looking at the frequency. However, we need to be cautious about it because they're a pathogenic and benign, both in DBSNP. One of the thoughts, though, is that if we're looking for a very rare disease and we find it with a high frequency in a population, it's probably not the cause. On the other hand, if we're looking for more common diseases, especially if it's recessive and we find it in the DBSNP in a fairly high population, that doesn't eliminate the possibility that it's pathogenic. We do look at prediction algorithms. However, this is more of a warm and fuzzy, and I haven't yet classified anything based on simply a prediction. I do do a literature search, and I'm not above Googling. However, my problem is that I don't know when to stop. If I can't find anything on it, is it because I've missed something or is there that there's nothing out there? Truthfully, I've taken up to maybe six, eight hours on a case, trying to gather this information, trying to put it together because I didn't know when I said, okay, I can't find anything, therefore, nothing's out there. This is a manual process, but we are now getting it into the fact, an automated process, which we are able to do checking our internal database, the local specific, DBSNP, those types of things. We are putting together in a pipeline so that we can come out with at least a smaller list of variants. So, as he had mentioned, there are different types of evidence that we look for, phenotype, genotype, cases with symptoms and normal controls. We look at functional studies, and that can be a wide variety of things that we take into. We will look at the splice predictors or amino acid severity conservation. What is important for us is we need to know if they've occurred with other causative mutations. And we need to put it together for recessive diseases as a combination. And we do, I do like genetic evidence and family concordant studies. These are not easy for clinical laboratories to get, however. And we have an IRB in place to be able to do that, but we've only done maybe a handful because the families just are not large enough to give us the information we would need. Now, the clinical lab can collect some evidences. It's not easy for us, but for example, we can attest some additional family members to show that a mutation is de novo. This is oftentimes in the pediatric population. The parents are very willing to work with us and get us the samples in. So we can show that a mutation is de novo. The linkage analysis larger families are more difficult. So we do go to prediction programs. And we've had a recent PhD, Bioinformaticus has been working on a project where he has worked on his own predictor using amino acid properties. But what has been useful for us is that he's combined it with other prediction programs. And in a sense, come up with a reference range which clinical laboratories like because we're used to working in reference ranges. But if you can see here by adding the scores up and he actually weights them differently that here are all of the known benign, the scores of known benigns, the scores of known pathogenic. This one uncertain falls right into there so it's predicted to be benign. Same way with this hair, this falls within the known pathogenic scores. One of the take home messages with this is that these predictions are much better and more accurate if they are gene centric and we don't try to have one set of parameters that'll cover every gene out there. A quick example, Alport syndrome because of a collagen, a defect in a collagen gene. You have an amino acid motif glycine XY. If that glycine has changed, it's going to disrupt the function. So for Alport syndrome, a glycine change is going to be much more significant than potentially a glycine in another gene. So when we try to collect additional evidence, we accept a very broad definition of functional. So for example, we will look at immunohistochemistry. Here's an example with the PMS2 gene for Lynch syndrome to see if it's expressed. So we can combine expression with a mutation that we're looking at. For biochemical diseases, we can look at either the enzyme or sometimes the pathway. For example, with MCAD deficiency, we look at acyl carnitines or we can look at transfer activity, sometimes structural analysis or going to an RNA. So these are the types of things that the clinical lab can collect. And I thought I would do just one brief case to show you what we're dealing with. So for example, this is in the PMS2 gene. We've found two different, I mean two different base changes at the same position. Now Ohio State has seven families and we've worked with them. ARUP, we have four families with this now. In Ohio State, there's been one biolilics. So if you have one mutation, then you're at risk for colon cancer with Lynch syndrome. If you have one mutation on each chromosome, so you have two mutations, it's actually a much more severe with brain tumors, et cetera, in a pediatric population. So we had a little bit more information that the PMS2 was absent by immunohistochemistry staining and it was a very unstable. So microsatellite instability was high. This is the best family that we could work with given this family. We had a 10 to one Bayesian factor or a likelihood factor that this is pathogenic. Obviously this would not fit to be classified as pathogenic if we used that alone. The amino acid predictions, polyphen, probably damaging P-mute benign, but with not very much reliability. Other ones leaning towards disease causing. It wasn't seen in 182 controlled chromosomes, which is not very useful when you're looking at very rare mutations, but it still can help us somewhat. The Western blot showed 50% protein compared to the control and haploid converted clones showed expression from only one allele. That starts giving us evidence, a stronger evidence that this is actually pathogenic. So my question, given that evidence, is that mutation pathogenic? Would you sign a report saying it is pathogenic? You're not helping me. So there is some things that the laboratory can do. We do have clients and we do get patients from many places and if we can collect some symptoms, some family history and a collection of previous lab results, we can start putting this together with the molecular results, the sequence variants, the known common polymorphisms and we are doing deletion duplication analysis with most of our tests so we can combine that information. And we do need to keep track of when we actually do have enough evidence to reclassify. Truthfully, we've done very little reclassification. The evidence to come in has been very slow. So what do we want? I want a clinically valid whole genome database. We need the variants, if the classification is known and this is probably going to have to be expertly curated, put it down. Again, if the severities are known, I'd like to start collecting that information. We need to know ethnicities and frequencies in different ethnic populations. We need to know the number of cases but it's much more difficult in a database if I see the same variant 10 times. As Les pointed out, they may be from the same family or they may be from different studies but I need to go in and look at each one of them differently. I'd like to have some type of a summary of the information. Obviously, we need to know the symptoms. We need to know if it's in conjunction with other evidence of other mutations. The evidence is one of the things I wanted to point out so that they're not all weighted equitably. The predictions, I weight very lightly. A case control study or a genetic linkage pedigree, I weight much more heavily. So we need to make sure that we're not putting everything, lumping everything together and the risks of an incorrect classification are different between different genes as well. There are sometimes I have to be much more cautious in calling something pathogenic. So the criteria for calling them pathogenic may be slightly different between different diseases. And because of that, I put this last is that as we're trying to do this, don't oversimplify. This is very complex. We have to take what evidence as we get. Those are not gonna be the same types of evidence for the different types of diseases. And the clinical labs are really very interested in helping and contributing but we have very few resources to do this and so any type of submission to a database based from the clinical side has to be very, very reasonable for us to do. And thank you. The university. I was wondering, we've talked about this before about how each laboratory is really working on this within their own kind of area of silo, creating their own databases and own resources of how to analyze these variants. Can you talk right now about the kind of collaborative efforts between the laboratories if they exist and also just on a day-to-day basis? Do you pick up the phone and call another director from another laboratory to try to work through some of this? The answer to the last question is yes, I do on occasion. Not often because we're getting enough data in-house but yes, I have picked up the phone and called another lab director. There actually is a collaboration going on. Donna will be speaking more about the ClinVar efforts and Heidi, Rem in the back, has really organized a large number of laboratories who have committed to be able to work together and submit the information they have to a central database. So there are definitely efforts moving towards that. You mentioned that you're working in an IRB to be able to interact with patients. It seems to me that, you know, what's made of it is welcoming talk where we're really crossing the line and really doing research on collection. And it seems like one of the things that would be important to discuss as part of this meeting is can there be creation of a safe harbor that would allow the exchange of information from the clinical side to the laboratory side in the context of trying to generate knowledge around these variants? Is this something, I'm sure it's something you've thought about but to have you in your discussions with others come up with mechanisms by which this might happen? We've had at length discussions with our IRB and they have basically told us that if we want to simply look at an individual for their mutation and not make it more generalizable, we can do that within the clinical realm. As soon as we want to say this mutation is now pathogenic and if we see it in family B, we want to call it pathogenic based on the evidence in family A that becomes human subject research. I don't agree with them to that extent and we've had discussions of what is research, what is clinical and at what point are they starting to step in into the clinical practice. But that right now is the definition and so if there is something that we can do to take back to the IRB and say these are some consensus that I can take to maybe loosen up some of these requirements, it would be very nice. So this is something that could potentially lend itself to say a conversation with the Office of Human Subjects Research which will guide us to say this is actually something that we would consider to be exempt. I'm assuming they would agree with that of course, but in terms of things to take away from this meeting that the NHGRI would potentially lead, it seems like this might be a reasonable conversation. Yeah, one of the problems is that for some reason, whenever our IRB sees genetics or genomics, all of a sudden it jumps it up to a high risk. Elaine, you mentioned sometimes thinking differently about a variant depending on the clinical context that you were having to evaluate that. For instance, when we have a prenatal case and we know a termination might be based upon our classification and so we're very cautious about that but it gets us into a dangerous situation where we're, I think, differently evaluating a variant's evidence depending on an outcome when I think our goal is to really be more standardized and there's so much literature out there that's incorrectly classified variants because of using non-standardized and not high threshold. So I wonder if you could comment a little bit more on that challenge. Yeah, so for example, we do a lot of beta-globin sequencing. A lot is known about the beta-globin sequence and there are other tests that we can use to tell them that they have something going on. So my molecular genetics test is not the one and only criteria they're going to use to take any action. It's only one piece of the puzzle. On the other hand, something like hereditary hemorrhagic telangiectasia that we have kind of a center of excellence there. A child may not have any symptoms at all but this could be potentially, this is life-threatening and so if we tell them that it's benign or likely benign, they're going to stop doing the intensive monitoring of that patient because this is the only bit of information that they're going on. So I guess in my mind it's a question of is this being used as the only source for an action or a lack of action or is it combined with other laboratory types tests that can also help guide? And so maybe that's in my mind. So for example, our genetic counselor with HHT said I want the greater than 99% probability for you to call it pathogenic. Well if I put that criteria to all of our reports in every disease we have, I probably wouldn't be reporting out many pathogenic mutations truthfully.