 All right, thanks. Hi. I'm Larry Bab. I'm a software engineer and I really appreciate the fact that you You guys invited me to speak at this conference and have this opportunity Have a lot to cover it's 20 minutes and I we all go through this when we get up here But I'm going to really try to go through it and get through it fast so we can have some some good discussions here Okay, my objective for for my talk is to really get into and demonstrate to the NHGRI group here that We really need to standardize variant specifications and baseline services to enable this technology and that we can do that in five to ten years and So we'll walk through and try to make that case With the hopes that we can realize that goal and get some increased investment. Oh, I'm not in slide mode yet Thank you perfect. All right software guys, right? All right It's all about us All right all right, so that's my my objective is again to Convey the message that we need richer Standardizations around variants and other things and baseline services in order to enable this technology and that it can be done in five To ten years. I think it can be done a lot sooner with the right level of investment But we'll go through that as I as I talk My experience here. I'm not gonna go through this in detail I just put it in there for information for those people that don't know me I don't haven't done this a lot. So I wanted to you know Make sure I gave some credibility to where I come from base essentially. I've been doing software engineering for 30 years 15 years in the clinical space and the last 10 years. I've been full-time all out either on the commercial academic and public side just contributing to as much as I can these these types of efforts around genomic knowledge Standards and providing IT resources to to help the community and the vendors out there Gain adoption to realize precision medicine goals. It is my passion and my calling at this point So Terry when she asked me to do this it came up with a title about bidirectional data flow from the clinic to the lab and back and It's great. You could give me any title and I pretty much was gonna talk about this So I'm gonna go through and really talk about it And talk about I'm not gonna get into the details here But I want to give the whirlwind tour of the state the current state of the clinic lab clinic workflow today It's pretty basic You know there's a lot of details behind these boxes but if we start in the upper left hand corner there and Go around from in a clockwise direction We can see that the essentially the physicians and the patients get together and at some point There's an order and they order requisition is sent to a lab along with the biosample. That's Almost completely done in paper form There are systems a lot of big institutions that have their their closed systems and have the money They they do have electronic ordering and that kind of thing But in a lot a lot of cases they don't especially if they're doing send-outs to labs They're not necessarily connected with but whether it's electronic or not It's still fairly crude and there's a lot of effort that goes into there and not a lot of information that's Captured necessarily that's needed by the labs But enough to get the job done the lab then goes ahead and runs the wet lab process to to do the assay and then that gets invested that gets set up and sent to their case repo where the pathologist geneticists genetic counselors fellows various people help draft the report analyze the significance of the variants that are found in isolation of the variants as well as in the context of the patient to develop a report if you're lucky you Have a lab that has a variant knowledge management system Most of them have Excel spreadsheets and all kinds of mechanisms to Sort of pull together their variant knowledge. You know some of them just rely a hundred percent on bioinformatics tools There's a whole mishmash of things that are done out there And then once that report is drafted. It is either electronically Sent back but most of the time faxed back to the ordering physicians facility at where it can be Uploaded into their EHR usually just as a narrative report a PDF or some electronic form Very little structured data ever makes it back into the EHR related to the content of the variants found the phenotypes and that kind of thing There's there's people in this in this room here that work at institutions where they invest a lot of money Trying to make this happen I happen to have the opportunity to work at partners health care for 12 13 years and They did some enormous amount of investment in making this happen and we did it And we did it with some some good success But it certainly certainly not something that scales to the wider community So we need the tools we need both the lab side software tools and the EHR vendor tools to really play in this space in order to make all this happen and I think was Bob Nussbaum yesterday said that that even if if the payers paid for this stuff and there was market Motivation there's still an enormous amount of labor that's needed to make this happen and we don't want to have a bunch of Different groups going through and building their own custom solutions because they just won't integrate It's possible to start lowering the technical barrier, and that's what we're going to get into a little bit more here Okay, I think job number one today The low-hanging fruit on the tree from an IT perspective is to get the variant phenotype standards Finally organized. We've been talking about it for at least I've been talking about it for 10 years now I've been working with a lot of different groups. There's a lot of good information out there If we get enough investment and the right group of people together. We can get the ball over the goal line and we're doing that with You know, we're making some progress in the clinical genomics working group I'm going to talk about that a little more detail But there's groups that are essential to this in my opinion today the ones that are really pushing on this are the Global Alliance and the ClinGen and NCBI EBI and the various groups that pretty much are participating with the Global Alliance All right, so this thanks to Bob Freeman for the next set of slides This is the idea here is this slides to demonstrate the time it takes to develop and the And how to generalize the interoperability starting from the lower left corner up to the right So we have academia consortia and SDOs and these are different kinds of use cases for standards where You have a shorter time to adoption because you have a smaller group and you're you're focused on a very specific problem Like you said your own little research study or a big research study and until you get to Collaborations where things get a little hairier and time starts to really go flipping by before you can get anything done And then finally the standards or organizations, which really are the end game to get everyone to adopt and those take the Largest amount of effort you can see here an idea of how that might play out This is not scientific. This is very opinionated, but but that's all right This I think it's right because it's my opinion and Bob's and we talked about it So, you know and and here's a sampling of the kind of Different groups out there that are doing this kind of work And here's just more of a concrete example of the ones that that both Bob and I are involved in that we believe Are are really the the resources and the projects that are going to make this happen I haven't seen any alternatives to this in terms of accomplishing the large-scale goals that we're trying to get to The clean gen data exchange in a Leo registry. I'm going to talk about that a little bit. Those are groups I'm highly invested in we're going to get into the Leo registry a little more specifically those tools exist today They're awesome and everyone should be using them The global Alliance work that we're doing is in in progress and they've just recently had a reboot where we've got this genomic knowledge Standards group that's that's working on formalizing computational representation of variant representations not just sequence variants but Copy number variants structural variants all those kinds of things with the idea that we can share that with the HL7s of the world and Get pilot implementations and real innovators out there driver projects to actually implement those things So that the fire groups and the HL7s can then formalize it and they can work on other problems other than Detailed genetic knowledge management and representation, which is a little bit outside their their Ability due to you know the level of volunteerism that would need to take to do that Okay So let's talk a little bit about HL7 HL7 right now is in a major shift to go from This version 2 version 3 sort of I don't want to call it arcane, but I just did so It's a sort of an older style way of developing. It's a it's a little hard to innovate with This new fire group has come along and and they've really done a complete shift where they're looking at increasing the innovation capability get more Innovators out there involved in developing apps having a little bit more control over it applying the standards and not giving such a high technical barrier to adoption You don't have to invest millions of dollars to to stand up an HL7 certified service You can actually go out there and build your use case for your example and get others to try it And if those things work then then eventually it will feed into the standards So it's really a very promising tool technically speaking or approach technically speaking, but it's still in the process of becoming normative It's not quite there yet The other thing to realize in the clinical genomic space is we are a moving target in that it's a rapidly evolving and complicated Domain and that that creates really conflating factor for developing standards, but we can again talk about that when we discuss things Here's some of the challenges just real quick is that the challenges around the clinical genomics work group Which I've participated in Bob's very involved in it There's a number of us Gil a whole bunch of us are are related to that group We go in and out, you know It's like joining a club and you know spinning and then you leave again You come back and and try to keep making progress and it does over time. It's just a little slow The reason I think is not because that's just because it's a dynamic industry and there's constantly evolving standards But it's just a broad and deep representation Domain in my opinion The clinical genomics domain is as big or bigger than all of the rest of hl7s domain put together And right now we have a clinical genomics work group of a dozen people that are really active maybe 20 It's gaining in in a membership But we just we just don't realize yet how that how to you know expand this to to get serious about Creating the entities and concepts we need in that space to to really get the standards over the hump It's it's like we're you know, I'm going off in a little tangent here. I but It's like we're given two or three Ways to try to solve all of genetics and we have to try to fit all this stuff into those two or three ways And it's very frustrating on the flip side the the hl7 fire gods and the people that sort of Decide when you can create these these large concepts and these standard concepts They don't want us to go out and sort of create this model that's not yet standard So it's this carton horse problem of show us the industry standards for these things and we'll give you the resources you need to extend and build and make the adoption really successful and Versus we need to build these concepts so that we can enable the technology to figure out what the standard is You know, so it's a carton horse thing and it's a real challenge So the clinical genomics work group Works really hard within the framework that they're provided by the fire group to build everything off these concepts called Observation and sequence a diagnostic report and that's it now that observation thing is really cool It's like a Swiss army knife of objects that you can do things with but you have to write a lot of Implementation guidance around it which is going to create a barrier for adoption Because people see that kind of stuff and they say this thing isn't ready for real time yet And I think that's a very big challenge. So how can we help? But the way we can help is to create standards in this space That take the burden off of groups like the clinical genomics work group and say here's how you do something That's that's really easy. I think it was Lisa yesterday when she gave her presentation she had the quote that I see codes are very portable. That's a perfect example Hgnc codes for genes. It's well adopted We don't have to have the hl7 work group go out there and design What's a gene? How do you deal with aliases changing things over time? People just use that now and they don't even question it ICD codes pretty much the same thing snow med codes same kind of thing their genome reference consortium with their with their Assemblies at five minutes. I have seven down here All right, great. I'm moving on all right next so So the the key to get there is in my opinion is work on these variant specifications and services. It's step one and What we need to do is we need to get investment in order to speed this process up. It's happening It's happening in a snail space. So until we get this done We're not going to get to the part where it starts really enabling this technology. I'm going to try to do VMC in One minute. This is the variant model modeling collaboration Reese Hart was in a group of us here In another places got together and started creating this this very specific Computable spec for how to define alleles haplotypes genotypes and that kind of thing and this has become the seed for what the global alliance is Is using to define standards right now? So really quick That's the group the mission here is to develop a reliable exchange of sequence variation in a ubiquitous way That can be adopted by all and that you know this gets into these things are not the same thing There's no way I'm going to get into Convincing the people in this room that there's a million ways to express the same variant and that's a very big problem You might think we're using VCF. We got it covered. That's not it. There's more to it than that and so So, you know, here's the example of some multiple representations What if we do nothing then we have this mess where you know Everybody's using different techniques for passing around this information in the end What we end up is with a lot of errors and challenges which really I think everyone understands and this Prevents people from investing in the IT needed to to get adoption happening all right 2015 we had a clean Jen had a collaboration with NCBI to create an allele registry Heidi Rehm got up at the I think it was the clean Jen decipher meeting down in Washington in 2015 and made the pitch That we really need this allele registry if we're going to figure out how to manage knowledge around variation It got EBI and NCBI so excited they came up right after the thing and they said we want to do this How do we do it? So ends up NCBI won that arm wrestling match EBI said go go for it We worked with Steve Sherry and Melissa Landrum and that whole group and they were fantastic work with him for almost a year and We got really far Steve Sherry did an amazing job of pitching it to his leadership there at NCBI And he was all on board and one of the biggest points that he brought out is NCBI doesn't want to make the policies for how variations done what NCBI wants to do is we'll stand up this public resource We'll make it ubiquitous so you can have a way for all these vendors out there to get reliable variation service to canonicalize and normalize variation in a consistent way to use in clinical care, but We can't be the policy-setting committee you have to have the community organization come together and do this and so we started to work on that we thought well Clint jen's a first place to start but we need to reach out and we got to get hl7 and global lines and all the people in This room on board of how to get that group together I'd like to consider this thing something like a variant Variation reference consortium like a VRC sort of like the GRC maybe so someone yesterday mentioned we should have a genome Medicine genomic medicine consortium to do this kind of stuff. That's fine however, it's done in the end NCBI ran out of money as all the government things pretty much do and And they were not able to support the full-blown picture and what ended up happening is they ended up creating this variation service and speedy Which ended up normalizing their dbSNP? ClinVar databases, I don't think it's done But in dbSNP2o you're going to see that a lot of these duplicate RSIDs and inconsistencies in dbSNP and ClinVar are going to get Resolved so there was some really good stuff that came out of it that is getting used by the and hopefully informed Global Alliance stuff so in the end then Baylor from the ClinGen did implement this registry They've done a fantastic job in this allele registry exists today And it's available and it's open to the public and it is fantastic from an IT perspective They did a great job of standing up a very performant high high throughput high access speedy Service and again, it's called the ClinGen allele registry. I recommend you going down to it It's at the link below what it essentially does is it canonicalizes variation. You can send it hgvs vcf you can upload Thousands and thousands of variants at once and if there's no ID in the system that it can already get it will automatically Register for you and return it and now you as an EHR vendor or a lab vendor You can pass around this ID instead of this big bucket of data that everyone does differently and try to have everyone understand The different representations you pass this ID and you have now a centralized service It says we all know this apple is an apple and it's not an orange regardless of how people put it in and it's just a really fantastic service these kinds of registries are being built today inside of people's Software probably in North Shore system and Intermountain system. I was talking to Mark Williams right before this What are they doing in my code for their knowledge management? He admitted we don't have that yet and we're working on it So we do it in e-merge. We do it in a lot of places, but this is hard stuff to do We don't want epic and sir. They're not gonna invest in doing this They're gonna wait for someone to do it. They may want to redo it their own way But but the the small guys and the innovators out there They can't even get started sharing variants until we do something like this I'm convinced and if we don't do it, we're just gonna be having this same discussion in 10 or 20 years So I'll leave that for the end can't go into all these slides But I'd refer you to go look at the slide deck. It's really interesting. Look at this service again Just giving example here's a whole array of things that you can pass in to search or bulk query or bulk download or upload There's over 650 million variants in there right now unique variants. This is not different versions of the same variant 650 they have all the exact clenvar when they put this data into this this repository They were actually able to validate the data that was in clenvar and they went back to clenvar and said hey By the way, you have some variants in there that don't seem right and they were able to work out fixing them And so this has really been a quite a collaborative effort between Clingent and NCBI and to improve the whole the whole Situation all right Here's an example of some of the resources that are now using this and integrating clenvar's putting these CIDs in there My varying info civic the consortium the VICC group the Griffith brothers from McDonald they have the civic Sematic variation tool they struggled with this when they set up that crowdsourcing thing They're like one of them, you know, there's a paper coming out I think in nature maybe next month that they're about to do and one of the sections in there is how they struggled with Normalizing the variation. I know I have 30 seconds But this is the hopefully maybe the last point So they they decided recently when they saw this they said, yeah, we're all in and they hooked into it now They're using it it doesn't solve all the problems But it solves a heck of a lot and it's a great building block to start with all right gonna go past that we need this trusted thing I sort of made that point. What's the next step beyond step one? We need the same kinds of things for phenotypes diseases and then we can put together and really start managing this variant knowledge broad in a broad way Not enough time for that obviously because I have three seconds four. Oh, I'm over That's going up not down. All right, and so we'll do that next time And thanks for everyone there and by the way on my slide deck There's some links at the end if you want to go check out some of the things I mentioned And that's it. Thank you clarifying questions. There will be an opportunity in a Little bit for the during the discussion section next up is says Chris shoot from from Hopkins I'm just gonna talk to spit about some of the harmonization around data syntax, so