Due to the rare occurrence of rare diseases and pediatric cancers, many of their biological details remain unclear. As a result, critical genetic links continue to go unrecognized due to a limited understanding of how certain inherited variants may impact the pathogenesis and progression of rare disease and cancer in children. This information gap not only presents challenges for patient education, but also for diagnosis and treatment plans.
Understanding the hereditary factors underlying rare disease and pediatric cancer equips patients and their families with information that is both educational and actionable. For instance, awareness of a predisposition allows patients to seek appropriate screening and prevention measures, as well as explore conversations surrounding family planning. Most importantly, these insights may aid in selecting the most effective treatment options.
In this roundtable discussion, experienced laboratory directors and clinicians will gather from a variety of institutions leading the charge in pediatric genomics research, including St. Jude Children’s Research Hospital, Children’s Hospital of Colorado, Nationwide Children’s Hospital, and Memorial Sloan Kettering Cancer Center.
Armed with a diverse breadth and depth of expertise in research and clinical diagnostics as well as genetic counseling and clinical care, speakers will share their perspectives on recent developments and current practices within the realm of clinical and research genomics. Additionally, they will describe how artificial intelligence and AI-based tools, including Genomenon’s Mastermind Genomic Search Engine, are streamlining patient care and forwarding the collective effort to advance genomic discovery for rare diseases and cancer in children.
- Describe key initiatives at selected Children’s and Research hospitals aimed at advancing rare disease diagnosis and care
- Understand how genomics is influencing the approach to treatment and management in pediatric rare disease and cancer
- Discuss the role of artificial intelligence technology to optimize diagnosis and healthcare delivery for those with rare disease and pediatric cancer
Amy Treece, MD
Pediatric and Molecular Pathologist, Medical Director of the Precision Diagnostics Laboratory
Children’s Hospital Colorado
Alisa Gaskell, PhD
Co-chair and Scientific Director of Precision Medicine
Children’s Hospital Colorado
Elise Fiala, MS, CGC
Senior Genetic Counselor
Memorial Sloan Kettering Cancer Center
Michael Walsh, MD
Geneticist & Pediatric Oncologist
Memorial Sloan Kettering Cancer Center
Roshini Sarah Abraham, PhD
Professor, Clinical Pathology, Founding Director, Diagnostic Immunology Laboratory, Associate Chief, Academic Affairs
Nationwide Children’s Hospital
Elizabeth Varga, MS, LGC
Director of Clinical Genomics Research and Development
Steve and Cindy Rasmussen Institute for Genomic Medicine
Nationwide Children’s Hospital
Kim Nichols, MD
Director of Cancer Predisposition Program
St. Jude Children’s Research Hospital
Jamie Maciaszek, PhD
Clinical Variant Scientist, Department of Pathology
St. Jude Children’s Research Hospital
Good morning, everyone, and welcome to the Children’s Hospital Genomics Roundtable! I’m Garrett Sheets, the content specialist for Genomenon, and I’ll be moderating today’s conversation.
Today, we are excited to host a highly experienced group of panelists from four leading pediatric hospitals, who will share their perspectives on recent developments and current practices within the realm of clinical and research genomics, as well as describe how AI-based tools, including Genomenon’s Mastermind genomic search engine, are streamlining patient care and forwarding the collective effort to advance genomic discovery. They have a lot of great information to share, so I’ll mention a small housekeeping item and then move right into our introductions so that we can get started.
We have a really full agenda today, and for those watching, we want to be able to take all of your questions. If we don’t have time to answer them at the end, we will send them to the panelists after the event and post them on our website. So, without any further ado, I’ll briefly introduce our speakers.
Amy Treece is a pediatric and molecular pathologist, and serves as the medical director of molecular diagnostics laboratory at Children’s Hospital Colorado.
Also from Children’s Colorado is Alisa Gaskell, who is the co-chair and scientific director of precision medicine, with expertise in the development and implementation of next generation sequencing technologies and the analysis of genomic data.
Elise Fiala is a senior genetic counselor at Memorial Sloan-Kettering Cancer Center with a focus on pediatric patient care.
We also have Michael Walsh from Memorial Sloan-Kettering, who is a pediatrician, geneticist, and hematologist-oncologist.
From Nationwide Children’s Hospital, we have Roshini Abraham, the associate chief of academic affairs and founding director of the diagnostic immunology laboratory in the department of pathology and laboratory medicine.
Also from Nationwide is Elizabeth Varga, the director of clinical genomics research and development at the Steve and Cindy Rasmussen Institute for Genomic Medicine.
From St. Jude Children’s Research Hospital is Kim Nichols, a pediatric oncologist and the director of the Cancer Predisposition Program. Her research interests include the molecular mechanisms that provide protection against viral infections and cancer.
Jamie Maciaszek is a clinical variant scientist in the department of pathology at St. Jude children’s hospital. A trained biomedical engineer, her research experience includes the molecular mechanisms of sickle cell disease and pediatric T-cell acute lymphoblastic leukemia.
Finally, Mark Kiel is the founder and CSO of Genomenon, where he supervises the scientific direction of the Mastermind suite of software tools. His research interests include stem cell biology, genomic profiling of hematopoietic malignancies, and clinical bioinformatics. Mark, I’ll open up the floor to you for a moment here. Did you want to say a few words to welcome our panelists?
MARK: Yes, sure, thank you Garrett! I want to welcome all of the panelists — I feel very privileged to have assembled so many folks from the clinical front lines — as well as to welcome our audience members. The very brief introductory remark I want to give is just to say that we at Genomenon really appreciate the time that you’re taking out to attend this discussion. I as a panelist hope to learn much more than I can share in this event. Without further ado, Garrett, I’ll pass it over to you so that we can maximize the conversation time.
GARRETT: Perfect. Thanks, Mark. What a stellar group we have! Now that everyone has been introduced, I want to start a question for the team from Children’s Colorado. Amy, maybe you can speak to this: In terms of genomic health care, what are some would you say are some key initiatives for your organization?
AMY: Yeah, sure! One of the important things to know about Children’s Hospital Colorado is that our approach to genomics currently is very clinically-oriented. One of the big things that our institution is working on is a precision medicine service line, and the goal of the service line is to combine clinical, laboratory, and operational expertise to really help make clinical genomics available to all disciplines. Now that we see genomics really expanding into a lot of subspecialties beyond genetics, we want to make sure that those specialties have the same access to genomics.
With this initiative, there’s a major emphasis on infrastructure to help support patient access to testing, access to counseling, and access to treatment. Our goal is to really have a consolidated approach to issues that cross specialties, so that we don’t see efforts duplicated in various different silos. Under this model, oncology, genetics, and other specialties are all at the same table, discussing their needs and identifying areas where they overlap, particularly in areas of really high priority for the institution. Though this approach is really clinical, the service lines will also focus on how to leverage our clinical data to support research initiatives as well.
GARRETT: Thank you, Amy, for that. Maybe someone from Nationwide — What are some comments on some key initiatives for your organization, and how do they relate to what was just shared?
LIZ: Yeah, I’ll take that one. Very similarly to the objectives at Children’s Colorado, we started the Institute for Genomic Medicine about five years ago at Nationwide Children’s, and the goal is to bring the research and clinical operations together under one umbrella. Instead of being siloed with clinical laboratory testing and a separate research operation, we have the ability to perform research genetic testing — such as whole exome sequencing, as well as some more research-based assays, like RNA sequencing, methylation testing and other specialty tests — within the operation of research, and also confirm that clinically, so that results can be delivered to patients across the institution. Very much like Amy and Children’s Colorado, we are trying to really expand our infrastructure so that it’s not siloed and increasingly incorporated into specialties. One example of that is Roshini operating in immunology, and bringing some of the next generation sequencing technologies into other areas of care.
GARRETT: Thank you, Liz. You mentioned Roshini’s work and immunology. Roshini, could give us kind of an overview about how your work relates to what has been shared?
ROSHINI: So my focus is on inborn errors of immunity. I came to Nationwide Children’s from Mayo Clinic. While I was at Mayo, I developed the targeted gene panels for the inborn errors of immunity, and what I rapidly realized is that we’ve now got over 450 different genetic disorders that fall under the umbrella of inborn errors of immunity. When I started the process several years ago at Mayo, I realized that, with the targeted panels, you need a sort of nimble approach; you need to stay up to date and keep your panels all updated. When I came to Nationwide Children’s, I decided that it would not be reasonable to recreate targeted panels for the inborn errors of immunity because there were commercial reference labs that were doing a far better job, and were probably able to be more nimble and stay current. So we now send targeted gene panels outside, but when it comes to exome or even genome under the research umbrella, then we look within our institution to sort of tease out complex patients with immunological disorders.
GARRETT: Thank you, Roshini. Now that we’ve talked about how these fields affect patients, I think I’d like to shift the conversation to general trends. So I’ll pose the question, and open this up to everyone: How has the approach to diagnosing rare disease and/or pediatric cancer changed in the past five years? How about someone from St. Jude Children’s?
KIM: Sure! Actually, I was going to mention this too in terms of our initiative, so I think this is a good segue. Let me talk briefly about what we have at St. Jude, and then I’ll just talk about how it’s changed. St. Jude, as many of you know, has had a really long history in terms of doing research on genomics and cancer through the Pediatric Cancer Genome Project. Six years ago, I was really fortunate to be recruited to St. Jude to bring on a clinical genomics initiative, where now we can offer all newly diagnosed cancer patients at St. Jude the opportunity to have their tumors evaluated using whole-genome, whole-exome, and RNA sequencing. We can also look at their germline using whole-genome and whole-exome.
That said, we are primarily a cancer institute, studying cancer and other disorders of the blood, so our focus is primarily on the cancer genes and cancer risk genes. But I think that what’s really wonderful about St. Jude is that, in addition to providing this now as a clinical service, the data that’s generated, as long as patients sign for research consent, both the tumor and the germline data gets uploaded through the St. Jude cloud, which is a tremendous research resource for institutions all over the world. I just want people to know that they can access germline and tumor genomic data, free of charge. I think, now, there are probably ten to twenty thousand genomes including both tumor and germline data in the cloud that people can access.
This has had a tremendous impact right on our ability to refine diagnosis of cancers. You can think about leukemia at St. Jude through Charles Mulligan now. What we used to look at under the microscope — I mean, I’ve been doing pediatric oncology for twenty years, and I remember when I started, you looked under the microscope (Mike probably knows this) and you see a lymphoblast, and a lymphoblast was a lymphoblast, and everybody got AOL therapy. Now, thanks to genomic advances, we know that there are multiple subtypes of leukemia, and some of them do better than others. We know we need to adjust therapies based on some of these molecular findings. From a tumor perspective, genomics and precision medicine has made a tremendous impact, particularly for leukemia, but we have a lot of work yet to do for many of the solid tumors, for example, and for some of the brain tumors. We’re just at the tip of the iceberg.
MARK: Kim, we lost your audio, so I’ll try to pick up where you left off. As a metapathologist by training, one of my professors spoke to the obsolescence of the microscope in a similar way to what you talked about. I don’t know if what you mentioned would have to do with its obsolescence, but if it would rather actually be the augmentation with molecular diagnoses. I was struck by how the WHO, the hematopathologic diagnoses in the WHO, were increasingly molecularly predicated. As I was leaving my training, I remember seeing that effect in solid tumors and brain tumors as well. I don’t know how much more that’s flourished since my training, but it was a very encouraging thing for me to see, where the diagnoses become much more molecularly predicated.
GARRETT: Really good discussion! I’m curious to hear, I don’t think we had any feedback from team members at Memorial Sloan-Kettering. Elise, Dr. Walsh, do you have anything to add on to build on what has been shared so far?
MICHAEL: No, I think I would agree with what the others have said. At least to chime in here as well, I think that we are, as Kim said, at the tip of the iceberg. I don’t necessarily foresee any other aspects of diagnosis going anywhere too soon because there’s so much variation of uncertainty that all of the additional information that we’ve relied on, historically, continues to be relevant in different ways. On the germline side, as opposed to maybe just the tumor side, that is probably even more true, as we’re dealing with so much variation of uncertain significance. Being able to phenotype and characterize tumors helps us better understand some of the predispositions as well, and spectrums that go with a lot of the cancer predisposing syndromes. I think we’re we’re still early days with a lot of this stuff. We have moved though, nicely, I think, from some of our original papers from five, six, seven or eight years ago, where we were describing a lot of the genetic predisposition, to more recently, how we’re actually translating this clinically.
AMY: I’d like to quickly add to that. I think something that I’ve seen change here in the past five years is that our genetic counselors on the oncology side are involved from the very beginning of diagnosis of oncology patients. They attend our tumor boards, and these discussions of predispositions happen right at the very beginning of diagnosis, not as an afterthought or a follow-up. So they’re utilizing molecular results as soon as possible to get those patients plugged into predisposition clinics.
KIM: Yeah, I was going to chime in, and this builds on what Mike said: In the old days, twenty years ago, a lot of guidance regarding germline genetic testing was based on basically family history or patient medical phenotype, but now we’re learning that a lot of families don’t have a positive family history, or patients with specific syndromes, Fanconi Anemia as an example, a good proportion of them don’t have clinical manifestations. So, having the ability to do genetic testing a little more broadly has really enhanced our ability to make germline diagnoses.
GARRETT: Okay! So we’ve had a really good conversation about germline variants, we’ve talked about trends that you all have seen in practice over the past five years. I’d like to now kind of look forward, and ask you — to pose again, generally, to everyone, because we had such a great conversation — how do you think this approach to diagnosing rare disease and pediatric cancer will affect things? Over the next five years, what do you see looking forward?
ALISA: I can take that one. Really, this builds on everything that we’ve discussed to date. I think what we will see is merging of fields and breakdown of the silos. It’s very comfortable to think of something as either due to an inherited germline or an acquired mutation, but I think the more we’re seeing and even moving into the transcriptional space, we’re learning so much, and we’re actually seeing that there’s more of a blending. Just as our tests are becoming unbiased, I think we should be approaching the diseases in a very unbiased fashion, and know that we have the expertise, and, based on the molecular profile, we can guide our patients through our system rather than following these very solid paths in the system.
ROSHINI: If I may add to that, in the inborn errors of immunity, I think a particular challenge is the number of new genetic defects we’re adding to the list. We’re at 450 or 460 now, and every two years, there’s an expert panel that gets together to classify all the inborn errors of immunity, and publish those classifications. In January 2020, the biannual classification was published, and then ten months later, we had to have an interim update because we’ve got about 30 new genes added. We can’t wait for another two years to update the classification. I think, for clinicians, it’s exceedingly hard to keep abreast of all these new genetic defects and the phenotypic variability.
Then, of course, clinicians are often faced with the conundrum of these variants of uncertain significance, and trying to find the correlation to the immunological phenotype or trying to do functional validation. In the immunology community, that’s something that we have to really work hard at, and we have a few thoughts and initiatives around that. How do we functionally validate these new variants or new genetic defects? How do we make sure that our community of clinical immunologists, but also hematologists, rheumatologists, anybody who sees patients with these inborn errors of immunity, how do we make sure that they are able to stay abreast of all this new information? So that’s the challenge for the next five years, I think, for us.
KIM: Yes, I absolutely agree. You know, Roshini, I think this goes across all conditions. As you think about it, it’s super exciting, as we discover new genes linked to various diseases, often those reports may be just one or two families that bring these these new genes to light. The problem then is that there’s very little published information about the genes and the genetic variants. So really, as new variants are identified, knowing if these are truly pathogenic or nonpathogenic is very challenging. It’s interesting. As we learn more, the number of questions is exponentially increasing, and we’re really going to have to work together as a community to standardize how we define changes, define conditions, or define phenotypes and put all of our information together to better understand what these new variants are going to mean. So I just wanted to second you very strongly. I think it’s a great time, but a very challenging time.
ELISE: I was just going to say something along the same lines. Better defining the outcomes of screening, now that we’ve brought in so much, has really altered who’s getting predisposition screening. The families who are told that they have LFS today are so different from the families who were told that they had it 15 years ago. It’s a much broader group, and that’s probably true for a lot of conditions, but LFS just stands out in that regard. We have the Toronto paper that showed the outcomes of that original screening, but I just don’t think that the population even looks the same as when they started that study. As the population is constantly evolving, we don’t know the outcome, and that’s only one condition that we really have a controlled study for. So I think that the more and more we’re all growing our surveillance clinics and the more and more people we have in them, the more we need to then publish the outcomes. We can know, did we even make a difference with all of this? What were the findings, what were the false positives, what were the true positives?
JAMIE: I was actually going to use the same exact example as Alisa did, with respect to TP53 and Li-Fraumeni Syndrome. With more and more unbiased testing and unbiased screening, we are seeing people in families that just have one cancer type; they don’t have the full LFS spectrum. In terms of surveillance, is the whole body MRI every six months, or every year really necessary? So there has been some conversation, which I think will inform diagnosis about expanding that phenotype, or having independent Li-Fraumeni and TP53 hereditary cancer phenotypes. As more and more people are have unbiased screening and the phenotypic spectrum expands, that’s going to also influence diagnosis and guidelines.
GARRETT: Mark, I think you were going to say something?
MARK: Yeah, I was just going to build on what Kim and Roshini had brought up, about the functional studies around confirming the pathogenicity of these variants. What Elise and Jamie had talked about, these larger-scale clinical screens and blending those two data sets together — My original question was going to be around functional screening across entire genes, and how much emphasis could be placed in taking a variant of uncertain significance with that functional evidence to suggest that there’s consequences at the protein level. Is there enough evidence from those studies to promote a variant from VUS to “likely pathogenic” within the ACMG framework? I just wonder, again, if it’s outside of a single, very focused paper about those variants; if it’s the result of the screen, is that evidence sufficient for diagnostic purposes? I’m posing this to the group, broadly.
KIM: I don’t have an answer, but I just wanted to say, I think that that’s a super important question. You really need to understand what the function of every specific protein is, right? And some proteins have multiple functions. There are these functional, or quote-unquote, “functional” assays that you can do to look at stability of message or transcriptional assay for a reporter assay for a transcription factor, but a lot of these proteins have multiple functions. I don’t know the answer, but it’s almost a philosophical question. If a transcription factor doesn’t function properly in a reporter assay, is that sufficient to call it a loss-of-function mutation? Honestly, I don’t know the answer. I would love to know what others in in the group think. It’s one piece of evidence but how much weight do you put on that evidence? Is that enough to convince you? And then, of course, there’s 20 odd thousand genes across the genome. We don’t know the functions of all of those. Developing one standardized functional assay that will apply to every gene in the genome is not going to happen. So how are we going to address this? This is something I’ve been contemplating since I moved to St. Jude, and I don’t have an answer, but I would accept any suggestions.
ROSHINI: I agree completely with Kim. Recently, there have been some papers on inborn errors of immunity predisposing to severe COVID disease. Without giving all the information away, because some of it is just about to be published, people have said, “X or Y or maybe 10 other genes — if you’ve got variants in those genes, they predispose to severe manifestations of COVID.” Then they cherry-pick and do certain biochemical analyses, but then they forget to do that with variants in the controls. Reading a paper that has all this information, you start to realize, “well, theoretically, they’ve done some functional validation of their variants, but because they haven’t done it in their controls, how are we to interpret that data?”
So there is one issue, the design of your population genetic studies, and then an additional issue of the functional validation. I agree with Kim. I mean, Kim all knows all too well the challenges with the inborn errors of immunity. We’re like mushrooms in wet grass, you know; we’re just sprouting new genes all the time. I don’t know the answer either. We’re using a hodgepodge of methods and so on, but as we’re getting to gene and variant curation, some of the gaps and holes are becoming glaringly apparent. We’re now finding ourselves soft pedaling or being more cautious about how we’re interpreting and trying to introduce in caveats. I think it is a challenge, because at least from the immune system perspective, we’ve just got too much variety to come up with any facile answers of how to do the functional validation.
MICHAEL: On the cancer side, it’s nice in the sense that we’re learning and being more creative about the way that we integrate the ohmic testing, from both the somatic, germline, transcriptome, methylation testing and trying to understand how we might be able to appreciate a second hit. Splicing might be better interpreted through different assays. On the cancer side there’s a lot of tools that we’re now availing of in a more meaningful way, as we’re getting beyond the initial steps of just describing frequency of variants and and honing in on certain genes in a more meaningful way as we think about them. The functional part will forever be a problem, both for economic reasons and for the way the medical system is set up in terms of our confidence of the rigor behind science, as well as for the mere difficulty of getting certain things published in a timely way. People will use what is being published as evidence, and usually, it’s not more than beyond a couple variants that you can hang your hat on from any given paper, in terms of your confidence for what something may or may not mean. I think that there are some societal or philosophical things that would would have to change gear.
MARK: If I could take that conversation a little bit further and expand it outside of the functional studies, one of the things that I’ve seen with respect to ACMG is more of a disease-specific focus, as opposed to the initially promulgated, very rigid, “this is the way you do the variant interpretation with ACMG,” and now you’re seeing a plethora of disease-specific focused modifications to the ACMG guidelines. Very similarly to what we talked about with respect to functional studies being predicated on disease specificity, I wonder if the group can speak to what changes we’re seeing to get more of a focus on disease-specific context for ACMG, and how that’s going to play out. When the goal of ACMG is to have a framework that can be abided by all, now we’re having these splinterings and ramifications, where it’s requiring increasing specialization within individual disease groups.
MIKE: Yeah, I’ll take a stab at that. In terms of the ACMG, they had a pretty daunting task when this all started. They were in a position where, in many ways, they started off backwards. They just started off with a gene list, and these were genes that you reported back before necessarily defining how you categorize things as “pathogenic” and “likely pathogenic.” It was a need born out of the sequencing that was becoming a mass-scale process. When one dissects the ACMG, you look to see who are representatives of a certain body, and who are medical geneticists that were part of that, and what their disease-specific areas were. You also have to look at what, in a medical genetics clinic, was defined as actionable, or at what areas where there were things needed to be reported back to management, whether it be for a connective tissue disease, or a surgery required for some of the extreme cancer phenotypes, or an arrhythmia might present itself.
As time went on, and that became the reference for people in other disciplines that were not typically card-carrying members of the ACMG to start with, there has been an initial acknowledgment that that was the appropriate place to start; that body has, through ClinGen and other efforts, integrated other colleges and societies. ASH is a perfect example of that, with Lucy Godley and her elite work leading the hematology efforts for classifying variants. An organic outcome of all of this is the additional experts that are being pulled in, and with that process, there’s been a whole lumping and splitting in terms of genes or disease, as well as getting in the more granular space of trying to define variants for all of these things through expert panels. So the ACMG started off with a very broad focus, but with that said, it was still somewhat narrow in terms of the representation. As time has gone by, there’s been a broader scope of folks that have been brought into that, and it’s served as a background for how all this variant curation has evolved.
KIM: I’d be curious, Jamie, what you think about the future as a variant scientist or curator. As it’s become more and more complex, how do you anticipate variant curators or scientists being able to analyze germline data? If every gene has its own rules, how will you keep up with all that?
JAMIE: Yeah, I think that’s going to be one of the biggest challenges. Within the ClinGen efforts, for most of our variant interpretation working groups, the collation of data from the reference labs who have all this internal data has been our most important piece of evidence for variant classification. Of course, as Mike said, some of our groups are using this functional data. In the TP53 working group, there are very well-established functional studies that we trust. In our CDH1 working group, we actually don’t take functional data into consideration, because it has not correlated with patient presentation. So really, the breadth of standardization across all of the genes in the genome to try to figure out what pieces of evidence are useful with respect to variant interpretation is based on the known patient phenotype. So that’s going to be the biggest challenge. The ClinGen efforts are focusing on the highest-profile, the most critical genes, LFS and CDH1, where there’s real surgical and/or surveillance implications. Expanding into those more moderate and lower-penetrance genes and/or variants is going to be very daunting. I’m not sure I answered Kim’s question, but it’s a challenge.
LIZ: If I could just chime in, too, as related to variant science. I think it’s been really interesting to observe the evolution of jobs in the area of genetics. As a genetic counselor, it’s really evolved. I mean, ten years ago, variant science as we know it today really didn’t exist, so I’m still surprised in talking to many of the labs that are hiring variant scientists. They are pulling from PhDs from genetic counselors, kind of converting them, and there’s not really even an organization that I know of for variant scientists in general. Also, with standardizing training, because it’s divided now among pathologists and geneticists — I don’t know if anyone’s really given thought, or Jamie, if that’s started to come up — What will be the standard for training of variant scientists?
JAMIE: I don’t think there is one yet. I mean, we have not discussed it at St. Jude, because our operation in terms of variant scientists is a very small, one to two person operation. But I do think that that’s very important, because reference labs and our children’s hospitals and everything do want standardization for what we are calling variants. I can interpret something completely differently than the person sitting next to me. Right now, I think that there is becoming more of a standardization. A lot of curators are involved with the ClinGen efforts, which definitely does help, having those conversations and seeing how these guidelines are formed. I think it will be important in the future to have a society, or some kind of more formalized training to get everybody on the same page.
MIKE: Yeah, I think that’s a nice point about the educational aspect of this, and almost hinting at a degree that could be had in all of that. When I got started in the space, so to speak, for three years, I just sat and looked at variants, and there was not necessarily any training. The ACMG rules around how to classify variants weren’t there. I had a genetics background from formal education, but it wasn’t specifically around molecular genetics. As this field has emerged, it’s still somewhat gray in terms of who knows what, and what people’s responsibilities are in terms of who should know what. To your point around variant scientists, there are certainly discrepant resumes, so to speak, in terms of who’s doing that. It is a big problem.
ELISE: The other thing parallel to that — The classification system is still “benign” to “pathogenic.” There’s not a formal recognition of moderate- or low-penetrance alleles. I just think that that will have to change at some point. Obviously, all these ClinGen efforts are a start, but it falls on people for the individual gene or the individual lab. Someone might have to write out a paragraph describing the effect of that variant, but that’s really so qualitative for something that should have more of a quantitative basis. We know that it’s so much more complex. Like you were talking about, Kim, all the different functions of a given gene, it is going to be different if all the functions are knocked out, or if one of the three functions is knocked out, it varies so much. It’s hard to have a variant classification system that only reflects a dichotomy of whether a variant is pathogenic or benign. I don’t have a great solution for how to modify that, because it’s so complicated, but it has to be more complex in the formal guidelines than just those two options.
MARK: Yeah, I’ve heard about a movement toward more quantitation of the scoring, as opposed to dichotomization. What I’m seeing in particular in terms of the disease-specific modifications is that there are ways to take these categories of ACMG evidence, even if they’re strong, and downgrade them to “moderate” or upgrade them to “very strong.” That does lend itself, as you said, to a more quantitative approach as opposed to just categorization, but that challenge is standardization. It makes it more complicated to ensure that Interpreter A is interpreting the same variant in the same ways as Interpreter B, even when they’re presented with the same information. I feel like that’s a very large challenge. We’ve talked about this a little bit, with respect to ClinGen and ClinVar. I’m wondering if we as a group can talk about any challenges that ClinGen and ClinVar are facing, or if there are ways that the community can get behind them to break through those challenges.
ALISA: As we’re looking at the challenge in front of us, maybe we need to step back and realize that we have this balance between continuous improvement and learning, and that is led by standardization, but then there are these individual breakthroughs. Every generation brings about 70 to 200 new variants, so to have a very structured, standardized model — the very subject matter that we’re trying to evaluate breaks that norm. So I think that while we need guidelines, you have to look at even mutation in the context of the person and the broader context of the genome. These are the kind of aspects where we as a community often forget that the clinical knowledge and what we’re seeing in front of us is going to be recorded into our EMR. How do we go back and learn from it? Once we realize how to harness that, and then build that into our continuous knowledge, that might be the way forward. Still has challenges
KIM: Yeah, you basically set the stage for exactly what I was going to say. I’m not as involved in ClinGen and ClinVar as I probably should be, I just have too many things going on, but the thing is, it’s really important that the genomic data be looked at in context with the clinical information. In my mind, one of the biggest challenges is bringing those two pieces of information together. You probably know about this, and this gets at what Elise said: There are some variants that are going to have a stronger phenotype than others, and phenotypes are going to vary. How are we going to bring it together?
For Jamie and the variant curator and scientist field, they probably come out from more of a pathology or computational angle. For understanding the clinical component, what’s really important is somehow having constant communication and a way to bring these disparate pieces of information together, so that we can better understand what these variants are going to mean in the future. If there’s an answer for how to combine and continually build on the clinical in a way that helps interpret the genomic, that’s what’s really going to be important moving forward. This is a really challenging but important area. Mike, I’m not sure if you want to build on that, as you’ve been so much more involved in this as a geneticist. I’m just a little old oncologist. What do you think?
MIKE: Yeah, the ClinGen part of it, where it suffers is — well, I shouldn’t say it suffers. I’m chair of the somatic germline integration committee to try to help clarify germline variants utilizing semantic data, and like anything, it’s resources. We all have very good ideas about how to try to take next steps, and then rate limiting steps are getting clean data to integrate and having access to information from different centers. With that said, there is good reason for certain centers to clean their data and have it organized at a particular level. That’s something I’ve appreciated more as time has gone by with the two most recent centers I’ve been with, in the sense that there’s this push. You free the data, free the data, free the data, but you don’t want dirty data. You want things annotated as well as humanly possible, whether that be the clinical part or the molecular part. Ideally, it’s both.
Sharon Plon, who leads ClinGen, she’s the godfather of all this stuff. That’s the way I think of her. She’s very conscientious of all this, and she’s been in this game a long time and has done a tremendous amount in terms of identifying players at different centers that can begin to really move all of this together in a meaningful way. When I mentioned the economic part of it earlier, a lot of it’s economics. Having the additional resources for bioinformatics and so on is always a great limiting step for all of this. Finding the MTAs to share the clean data when it’s available can be a rather drawn out process. Those are just a couple practical things, but they’re things that can actually expedite matters quite a bit, if in place.
LIZ: Mark, I was wondering if you could comment on this. I was just thinking, and I know this is kind of a topic we wanted to discuss, about the role of AI and machine learning here. That’s what I’m hearing a lot of when we’re thinking about the clinical correlation and design of products that can really use the machine learning concept to pull that clinical-phenotypic information together with the molecular. Mark, do you have thoughts, or does anyone else have thoughts on initiatives at their own centers to do that?
MARK: Yeah, I’ll start by saying that, as clinicians, you should always be be armed with evidence. Genomenon does not think of AI as a black box solution to any of these problems. I think that AI is a tool to arm the clinician with the best evidence, and to continue to leave the decision-making in the hands of trained clinicians for all of these nuances that we talked about. In a clinical context, the challenge of these disease-specific diagnoses is just the nature of the complexity of the data. I think of AI as a way to organize evidence for decision support. I don’t know that we, as a community, would ever want to get to a place where AI could be a stand-in for any of these clinical decisions.
Particularly, at Genomenon, we surface information from the medical literature and keep that as up to date as possible, but the idea is always that it be as sensitive as possible and as organized as it can be. Then the ultimate onus of decision making is on the clinicians. I wonder if anybody else is more optimistic about AI taking over? Personally, I’m not. During my training in pathology, I was always very pessimistic about the idea that these tools would replace the human touch.
AMY: I don’t know if that I have the answer to that, but I think one thing this brings up for me is that it’s important to think about how we enter our molecular data into the electronic medical record. With our most recent assays, we were actually able to interface our results directly into the EMR. We now have discrete variants that are searchable, which for us was a huge victory, because now we can actually tie that data to particular patients and particular demographics, but that’s still not perfect either. That’s something we really need to think about — how do we make that better? How do we enter our molecular data into the electronic medical record in a way that makes it useful for all kinds of different things, for interpretation, for submitting data to consortia, for utilizing that data to talk to payers about putting together lists of patients where one test was really impactful? Organizing our data and interfacing that with the electronic medical record is going to be one of the first steps to making AI and decision tools really valuable.
ALISA: If I could add to that — I think, Liz, you mentioned that we need to evaluate the role of clinical labs in this journey and we have to recognize that we’re not at the end. So while we’re entering and interfacing the genomic data that was born out of a subset. I think we mentioned very early on that, as our knowledge is growing, rather than having these very discreet panels, we know that we’re generating a much larger data set. We have the pressures of the payers, and we cannot do exploratory panel research. The lab is actually generating that data. So how do you link that larger body of data with the EMR if you have what I call an unsatisfactory diagnosis? There are boosters in there, can we go back and ping that data in a very streamlined fashion, rather than just pump the brakes, let’s sit down, pull this data set? Can this happen in a more streamlined fashion?
MARK: Alisa, you brought up something that I wanted to touch on. With all this proliferation of new information with modifications to the decision making and these VUS’s in otherwise undiagnosed patients, how does the panel think about the look back challenge, if you have a case that remained undiagnosed? Michael, you’re bringing up this pragmatic challenge, trying to figure out who’s got time for this, who’s got the resources, let alone, how do we communicate this information to the patient? If anybody has ideas about what you’re doing at your institutions to solve this look back challenge of cases that haven’t been diagnosed, but demand a revisitation of that data in light of any published information?
KIM: From our perspective at St. Jude, it really behooves the provider to go back to the lab to ask for a reanalysis. I’m not sure how it works at other places. I’m worried that from a laboratory perspective, it’s going to be too challenging or overwhelming to go back and look at every single case that’s been tested over the prior years. There will be hundreds to thousands of cases. At our institution, it’s been up to the genetic counselor or the provider to ask for review analysis every couple of years as needed, and lo and behold, we’ve come up with new diagnoses as new data has come on board. I’m interested to know if others feel differently, but it’s worked for us. You just have to educate the providers that they will need to reach back out. The question I guess will be economics. I don’t know if it is something that the provider would need to pay for, or if that can be done at a reduced cost or free for service. I think for our patients, the companies that provided the broader sequencing at the time did the re-analysis either for free or for a reduced cost, but I don’t recall. I’d be curious to know what others do or suggest in these cases.
LIZ: One thing that I found helpful when working with Genomenon and interfacing with labs was just learning about their different operations. When you think about some of these tertiary analytic platforms that are now prioritizing variants, as well as other AI-based tools with updated literature, I’m wondering how much can be more AI-based for the look back, or at least presented to the clinician? Like, telling the clinician, “here’s what’s been updated over the last year, here’s the new information about this variant that’s been reported,” which then helps the clinician to triage that look back process. I just think the only way to do it is some kind of automated supplement to that process that preserves the hand element of discernment. I haven’t really come across anyone who’s optimized that process. I’m curious to know if anybody else is aware.
ALISA: I also think that there is a power of the patient. The patient has a voice here as well, and there’s a huge educational component. When we go into surgery, it’s very finite. We know exactly what it is, whereas when they’re entering and getting the genomic testing, this is a journey, and they have to understand that today’s result doesn’t mean that it’s going to be the same tomorrow. Having discussions on multiple levels will help guide us to what that process looks like. As always, just like with our iPhones, some people will say, “track me, I don’t care,” and some people will make sure that everything is well-guarded and put the aluminium hat on their head. So I think we have to be very sensitive to those privacy aspects as well.
ELISE: Yeah, I don’t think we’ve optimized it, either, but I think it is very patient-dependent. The way that we have it, or the way that most people have it, is that the patient has to reach back out to us for clinical updates. If it’s part of a research protocol, like our MSK impact assay, then there could potentially be updates, but only if that patient is signed on to the research protocol. If we find something and then are not able to re-contact the patient, if we can’t track them down and the phones are not in use, then we have a protocol for how to deal with that, but I think everyone’s really nervous to take that responsibility on clinically in addition to the huge workload. It’s just a lot of liability. It’s not a perfect world, where you can find everyone who you could find two and a half years ago.
It is very patient-driven, and that also allows patients to kind of dictate the importance of it, because it’s just not worth our work if they’re not interested or they’re not going to follow up regardless of what we contact them with. There are some patients who really have something, and we know they have something, and they know they have something and we’re all frustrated that we don’t get to it with the first test. Then there are other patients who we send the same testing on, but they probably don’t have a predisposition, and the negative or unlikely VUS’s are reassuring to them. It’s probably fine if those patients don’t recontact us or if they reach out in five years instead of one or two. So we just try to emphasize it to all patients, but we’re not doing anything clinically on our end without them reaching out.
MARK: I wonder in the last few moments if we can build on what Alisa and Elise said about patients and and their intrinsic motivations and their understanding. I’m just wondering how well-educated they are. In light of direct-to-consumer genomics, how receptive are they to these new lines of evidence, and how has that influenced your clinical practice? Or has it not changed at all?
LIZ: As Elise was talking, one thing I was thinking about is health literacy barriers and diversity and access barriers, because it’s incredibly disparate. That’s also a concern of mine, that if it is left to the patient, even though that can be very positive, due to barriers that some patients have over others, that could actually lead to a lot of inequity.
ELISE: Yeah, completely.
KIM: That is obviously a problem, but you know, Liz, I would imagine, even people who don’t necessarily have those barriers could have problems. I don’t know as much about direct-to-consumer testing, but some of those tests are for very specific variants in a gene. If a patient gets a negative test and they think that they don’t have any variants in the gene, they think they’re okay. This just emphasizes, for all you genetic counselors, how critical it is to do genetic testing in the context of genetic counseling. Personally, I feel like it’s a real disservice to do genetic testing without genetic counseling on both pre-test counseling and then post-test counseling. We’re lucky in pediatrics and at St. Jude. Because our patients have cancer, we can see them periodically afterwards to continue to educate the families, but it’s very different perhaps with other diseases and in the real world. The counseling is really critical across the board.
MIKE: Yeah, you often find that providers don’t have a clue about a lot of this, never mind, people trying to explain it to patients.
GROUP: (LAUGHTER, AGREEMENT)
GARRETT: I just want to thank all of our panelists for taking the time to offer your expertise in this roundtable. I think we had a fabulous conversation! It seemed to move itself. You’ve all had a lot of great feedback, and you had some really good conversation chemistry, so I think it was very valuable. I thank you for that. I will close out with this ending slide here.
For everyone watching, thank you for tuning in, and to all of our panelists, thank you for taking the time to offer both your expertise and your time in this roundtable! If you have any questions or would like to discuss your variant interpretation needs, feel free to send an email to our folks at email@example.com. Additionally, if you don’t already have a Mastermind account, you can create one for free and receive a free trial of Mastermind Pro. Last but not least, your feedback is very important to us, and we’d love to hear your thoughts on today’s webinar. At the conclusion of this event, you’ll be prompted to take a brief survey that will help us as we continue to create a highly valuable customer experience.
With that, thank you so much for your contributions and interactions, everyone! I think we had a great conversation.