HL7 is holding its annual conference on genomics February 20-21 in Washington, D.C. The theme this year is focused on the intersection of clinical genomics and artificial intelligence. To preview the meeting, I recently interviewed Grant Wood, a member of the HL7 Clinical Genomics Work Group and a senior strategist at Intermountain Healthcare’s Clinical Genetics Institute, who is chairing the meeting.
HCI: Why did HL7 choose to focus on the confluence of clinical genomics and AI this year?
Wood: AI and machine learning are getting a lot of hype now. We wanted to bring together people involved in patient care delivery, lab testing, and developing data standards to look at where AI and machine learning could mix with clinical genomics, and what the issues are that we would have to overcome to achieve that.
HCI: I understand you did an informal survey of your contacts in the field to get an idea of areas of promise. What did you hear back?
Wood: A lot of the responses were about how machine learning and AI could help with clinical work flow. Those are the types of practical ideas I was hoping to get. We weren’t looking for a pie in the sky where AI is going to crunch through a bunch of large data sets and all of a sudden we are going to come out with greatly expanded knowledge around what each individual’s genomics mean. But we knew there were opportunities to improve or augment the clinical genomics process.
People talked about bringing datasets together that haven’t been brought together in the past and being able to analyze those combined data sets. From that we could develop new apps or tools for clinicians to use because they have more insightful knowledge around the patient. We have to bring together these data sets that have previously been siloed. These could be complicated data sets. This is where we think machine learning could add value because now we can bring them together and come up with new insights and incorporate those insights into the work flow.
HCI: At the HL7 Genomics conference a few years ago, participants bemoaned the state of interoperability between family history tools and EHRs and the limited capability of most EHRs to ingest structured family health information. Is that changing?
Wood: Not as much as we would like. As we talk about combining genomics and AI, there is a realization that the interoperability piece has not been completely solved. But maybe with the excitement that comes from AI, we might double or triple our efforts to solve the interoperability issues. With so much money put into AI research and development, it might be one of those external drivers to solve the interoperability issue. I am always trying to see how different drivers might emerge to solve the problem. And this might be one.
HCI: On the standards front, there have been some promising developments with FHIR Genomics and Sync for Genes. What is going on with those?
Wood: Right now we are going through the final process of developing a FHIR Genomics profile through HL7 that we hope to bring to the normative ballot in the fall of this year. We also are trying to promote and ramp up opportunities to continue to test the FHIR Profile. Sync for Genes is one of those activities. At a recent FHIR Connectathon, people were using FHIR Genomics profile to develop new apps that are doing interesting things. The plan is to have all these things going, so that when we have that normative standard, we can tell the world it is time to adopt it.
HCI: I see that Peter Goodhand, director of the Global Alliance for Genomics and Health, is going to speak at the HL7 meeting. That group has unveiled a strategic roadmap involving more than two dozen projects to be launched in 2018 and developed over the next three years, laying the groundwork for real-world genomic data sharing by 2022. How significant is that work and is HL7 involved in those projects?
Wood: HL7 is very deeply involved. The great thing that the Global Alliance is doing is trying to solve the issue of the sharing of data. They are creating a federated model, whereby we can expand the data sets that people are using to do genomic research. You could just imagine as they are solving that issue, the development of machine learning algorithms for use on that federated network of data. The promise of new discoveries coming from that is pretty exciting.
HCI: On your meeting agenda, there is a wrap-up session on creating a new grassroots coalition that will develop a plan to promote an action agenda on clinical genomics and AI. That sounds pretty ambitious.
Wood: We are asking attendees to bring their laptops and tablets, and we are going to be working on the same Google Docs, typing in ideas on specific topics. We plan to use that input to compose an action agenda. The plan is to have that document ready in advance of the HIMSS conference so we can distribute it and contribute to the conversation. We also want to ask key stakeholders about the idea of the development of a coalition. If it emerges there is value in that idea, HL7 will try to support that effort.
