FHIR Chat · GA4GH vs HL7

Dear all,
My background is genetics, with no knowledge in genetic data exchange. However, I am trying to read up on ways to facilitate genomic data exchange, especially those supported by FHIR. However, I struggle a bit with distinguishing HL7 CG from GA4GH GKS. I understand that these collaborate to optimize exchange of electronic knowledge, but I don`t understand their different roles. Could any of you please give me an elementary update on this? Do these two groups offer different platforms, and if so, are they both in linkage to HL7?
My understanding is that GA4GH is involved in standards, like standard API, standard reference sequence etc. Whereas HL7 CG are more involved in projects regarding the technical data exchange. Is this correct?
I am sorry for my lack of knowledge and questions regarding this theme, but I do hope some of you can find the time to try to explain.

Thank you and kind regards

Bob Freimuth (Oct 30 2019 at 15:36):

Hi Eva,

This is a good question. The standards dev space can get crowded with overlapping, complementary, and sometimes competing efforts. I am co-leading efforts in both HL7 CG and GA4GH GKS. One of my roles is to try to keep the two groups aligned, which can be a challenge. I'll try to answer your question briefly here, but if you'd like to follow up we can add this as a topic on a future CG call.

HL7 CG is focused on clinical genomic results: those that would be obtained by a clinical lab and reported to an EHR for use in clinical decision-making. Therefore, we spend a lot of time thinking about the structure and content of the lab reports that report genomics data (of all types), and how we can develop standards to support the clinical use of those data. HL7 is ANSI-certified and its standards (e.g., v2, FHIR) are often mandated for use within clinical systems (e.g., EHRs).

GA4GH GKS has roots more on the research side of the translational spectrum. It tends to develop a broader array of standards, formats, policies, etc that support the exchange of genomic data among a variety of research and clinical systems. The products are less rigorously developed, as the organization is not ANSI-certified, but they are targeted to real-world use cases and are intended to be rapidly developed and implemented. The scope of work in GA4GH is broader than HL7, and includes security technologies, regulatory/ethics issues, cloud-based workflows, data/knowledge representation for genomics and related clinical data, and file formats (e.g., VCF). Its standards are implemented by large national initiatives and small research labs.

Both organizations use consensus-based development among a relatively small number of active contributors. HL7 strives to generalize for broad adoption from the start, whereas GA4GH's products are linked tightly to "driver projects" that guide development and commit to immediate adoption (making those products a bit more narrowly scoped initially). Culturally, HL7 tends to be more heavy-weight and process-driven, while GA4GH is more agile.

I believe there is need for, and room for, both organizations in the genomics standards space, as they have different immediate goals and support a slightly different set of end users as primary consumers. The key is for all SDOs in this space to work together to develop standards that are interoperable - such that the semantics of core elements are shared and the syntactic differences are straightforward transformations.

I hope that helps. I've given several presentations on this topic and I'm happy to discuss further.

Allison Heath (Nov 01 2019 at 15:58):

Thanks - this was helpful as I was curious about this topic as well. We've been more on the translational research side (and thus GA4GH), but there's been increasing interest in how to flow data from (and maybe ultimately back to) clinical workflows. So have been interested in understanding the HL7 CG side a bit more, especially as we likely have some practical examples over the next year where the standards need to meet/align...

EvaM (Nov 29 2019 at 12:35):

Thank you so much for your answer Bob, that helped a lot on my understanding.
I would really like to discuss this further, as I have shifted from a research background to a governmental position, where I will try to aid both the clinical and research community in my country on the possibilities to exchange genetic data. FHIR is a recommended standard in my country, and if I understand it correct then genetic data can be exchanged just as easy through GA4GH as with the FHIR API?
Security is always an issue, and I understand that the Beacon API has taken that into account. Is that also a fact for FHIR? And is the Beacon API an open API that can be easily employed by researchers (and clinicians) in all countries?

You write that HL7 CG is focused on clinical genomics results, whereas GA4GH has a broader perspective and roots more on the research site. However, in a clinical setting there is often a need to exchange genetic information between researchers and clinicians. I.e. for all clinical studies. Would it then be more optimal to employ the GA4GH API for all parts, or can the APIs from GA4GH and FHIR work together?

I am sorry for my lack of technical understanding in this field, but I greatly welcome all information that can help me broaden my horizon and understanding.
Please add this topic as a future CG call, and I would really like to follow it!

Thank you and hope to hear from you soon!

Bret H (Nov 29 2019 at 15:53):

Firstly, it is not G4GH vs HL7. The two have members co-participating in both efforts. Is there anything stopping us from formalizing the relationship? Many would state that the constructs used in the HL7 CG WG guide to transmit variant level details is adequate for research needs with regards to variants. Yes, bioinformatics is currently needed to normalize the various ways that genetic variant level data is reported. @Bob Freimuth is a good goto person.

Bret H (Nov 29 2019 at 15:54):

Secondly, think carefully about which part of a Clinical Genomics report is needed for research. I have seen research projects that need the highest level (specific biomarker presence) to the lowest level (position, reference and change).