At the 2019 American Society of Human Genetics (ASHG) Annual Meeting, Dr. Kasia Ellsworth of Rady Children’s Institute for Genomic Medicine presented at the Genomenon CoLab Session:
Automated Literature Curation and Variant Interpretation Pipelines with Mastermind following rWGS within the Pediatric ICU – the RCIGM Experience
About the Speaker:
Dr. Kasia Ellsworth is Associate Lab Director at Rady Children’s Institute for Genomic Medicine. She’s completed fellowships in both clinical biochemical genetics and clinical molecular genetics like the Greenwood Genetic Center in Greenwood, South Carolina, and is certified by the American Board of Medical Genetics and Genomics in both specialties. She holds her PhD in molecular pharmacology and experimental therapeutics at Mayo Clinic College of Medicine in Rochester, Minnesota.
A Transcript of the Talk:
Thank you so much for the introduction. Today, I would like to focus on how we, Rady Children’s Institute for Genomic Medicine, are utilizing tools such as Mastermind in automating our variant curation in the very unique setting in pediatric ICU.
Rady Children’s Institute for Genomic Medicine is actually embedded within the Rady Children’s Hospital, which is the largest hospital in California. We began our genome sequencing in July of 2016, and are currently a CLIA certified and CAP accredited laboratory. You can see that our mission is to enable the prevention, diagnosis, and treatment of childhood diseases through genomic and systems medicine research, with the goal to make genetic testing fast, easy, and routine care for diagnosing and delivering precision medical care to critically ill babies and children. Our focus is on the intensive care unit in the pediatric setting, where many of the sick children end up. This setting provides a very unique opportunity to implement precision medicine, and to have significant impact impact on the lives of children.
Roughly about 10% of children end up with severe chronic illness. However, that number accounts for about 70% of the hospital costs. Thus, to understand this population of patients is crucial to not only benefit the patient’s themselves, but also for hospital economics.
So here is illustrated our current standard of care. You can see that roughly about 4% of children or neonates born in community hospitals end up in the NICU or PICUs so they require critical care. And while they’re admitted within this intensive care unit, interim empirical treatment as well as search for etiological diagnosis will start, and those babies can then improve or worsen, but they can get better and be discharged home, or there will be a palliative care decision made. And unfortunately, some of those children end up dying. However, if we are able to improve their condition and modify the treatment, the outcome can be very different.
This is how we envision the future of the ICU care units where critically ill children are admitted to the hospital. While the search for an etiological diagnosis and empirical treatment are underway, about 40% of those children of the 4% that are admitted to the intensive care units would be eligible for a rapid whole genome sequencing. Based on the data that we gathered so far, we expect that about 20 to 50% of those kids will end up having a genetic diagnosis, followed by a higher percentage of those children where diagnosis will inform precision treatment, leading to avoidance of mortality and long-term morbidity and shorter clinical stay with decreased hospital costs. And our immediate turnaround time to deliver the genetic diagnosis from sample accession to preliminary diagnosis is around three days.
Here’s an overview of our process that we perform today at our institution. We collaborate with more than 20 children’s hospital nationwide where consent and enrollment takes place. At this time, we also receive patient phenotypic information. Once the sample is received and DNA is isolated, sequencing takes place. We perform sequencing on lllumina’s NovaSeq, and then alignment and variant callings being made for SNVs and indels via Dragen, and for the copy number variants via DNA Nexus and our in-house developed pipeline. Then the combined VCF files are annotated in Fabric, where filtering and interpretation and reporting takes place. Once the report is generated in Fabric, it’s uploaded into the electronic medical records. This is a clinically validated rapid whole genome sequencing correct workflow. And if any of the variants require confirmatory testing, we perform that in-house as well. Again, our median turnaround time is about three days from the sample accession to the generation of translated results that is being uploaded into medical records.
Now I want to focus on the case analysis. You can see that there are three major hubs that are part of the case analysis, including the variant prioritization where ranking of the variants takes place in our interpretation software that we are using. Next, variant curation, where we research variants for evidence of pathogenicity. Then if we decide to report any of the variants that we’ve identified once they’re curated and all the evidence is gathered, then we classify them according to ACMG criteria. But this is really an area where we spend a lot of time, and truly by improving and automating some of the variant curation parts, we are able to contribute to a decrease in turnaround time, and that’s really where Mastermind comes in place.
You can see as Dan mentioned in his talk (Dan Bellissimo of University of Pittsburgh Medical Center. Watch the recording) that there are various databases and tools that we use for variant curation, and one of them now incorporated in our workflow is Mastermind. To illustrate that, I would like to share with you this one case where truly being able to automate the process and decrease the turnaround time to deliver the final results has made a huge difference.
This is an ultra-rapid case where fast turnaround time really does matter. There was an eight day old baby boy who presented due to a newborn screen being positive for SCID, and was admitted to the NICU. It was a duo analysis, as the baby was born from egg donor via IVF, and when we received the sample into our laboratory the flow cytometry confirmed in this baby sample the absence of T and B cells, so therefore his SCID was classified as T negative, B negative, and NK positive SCID.
This is a screenshot of our analysis software, and you can see when we started the analysis we input what the HPO term is for the main phenotype associated with the patient’s presentation, and then while we started curating the variant (this is a particular snapshot looking at the copy number variants) we identified this large deletion and three exons within the gene called DCLRE1C. So then we’re curious, what is this gene all about? When we look into OMIM then, we could see that this gene is associated with autosomal recessive SCID, and an estimated 4% of severe combined immunodeficiencies are caused by pathogenic changes in this particular gene.
Based on what we gathered so far, based on the patient’s phenotype, the match was really really strong. In addition to having absence of B and T cells, this baby also had a chest x-ray showing absence of thymus, which was really consistent with this disease. Infants with this disease usually present a little bit later in life with disease presentation being due to severe recurrent viral bacterial or fungal infections and failure to thrive.
Our baby was doing really well in the NICU, so we got him prior to severe presentation. When we look further in some of the databases, we could see that the deletion that we identified was previously reported in affected individuals, therefore the classification for that was pathogenic. However as I mentioned, this is a SCID that is autosomal recessive. We continued analysis of the genome and we identified another hit in this gene. There was a missense variant, therefore we could see that it was not previously reported. It was absent from the population, and in this case because it was duo analysis, was inherited from the father, the deletion was not.
So next step when we perform the case analysis and variant curation, we open different sources that we have available. In silico analysis in this case for this particular missense case was pretty strong. Then we look at the analytical validity of this call. You can see it’s a little bit of a skewed call in this case, and this is actually due to the fact that this particular gene has the pseudo gene that is homologous to the regions of exon four and six to nine, and this particular variant is within exon 6. Therefore, the skewing is most likely due to the pseudo-gene presence. Most importantly, what we actually learned from our immunologist was that this gene is not really difficult to sequence, and variants could potentially be missed or filtered out, but we got it.
Then we look first into HGMD to see whether or not it was previously reported, and it was absent. So then we continue the curation of the variant. You can see that actually we have in our variant curation workflow as well as in our software that we analyzed the cases, this particular link out with the letter M. This is a link out to Mastermind, and this variant is present in Mastermind, and means that it has some publications associated with it. So this window pops up when you click on it, and you can see that indeed there’s one publication that talks about this particular amino acid position.
When we opened the article in Mastermind, we could quickly access and read the data, and it showed that in vivo and in vitro studies were performed and this variant’s name was actually under the protein’s name which is Artemis, rather than the full gene name which is different. What we learned from reading this paper is that this particular amino acid position is actually one of the amino acid positions that are essential for the protein catalytic activities, and therefore we were able to evoke functional evidence.
Both variants were then confirmed and reported as likely pathogenic and pathogenic we could not obviously provide the pasing of this variant as we did not have a maternal sample available, however one of them was detected in the dad and the other was not. Based on this clinical presentation of the patient and molecular findings, the clinical team actually agreed that this is indeed a patient’s diagnosis, and very likely it was done prior to onset of severe symptoms. Moreover, this particular type of SCID is difficult to transplant by a conventional BMT, and because the baby was diagnosed quickly and prior to onset of symptoms, he was actually eligible to be enrolled in the gene therapy trial, and so far is doing great post-transplant of the genetically corrected cells.
This is just an email we received from the clinician taking care of the patient saying that “Knowing the genetics rapidly made this gene therapy available and able to be performed prior to any infections”. So, that was really stunning results, and the strength of Mastermind of providing us additional information for variant classification in a timely manner.
This table summarizes some of the studies that we performed in other cases. This is a rather large table but you can see that for several cases and the subset of variants that we looked at, the number of references from the Mastermind were very high as opposed to HGMD references. Moreover, a subset of those references were very useful for our curation, and what is very important for many of these particular variants is those were diagnostic variants that were reported in the patients reports.
In summary, I showed that there’s clinical utility of performing rapid whole genome sequencing tests in the pediatric ICU. Specifically that rWGS can help to guide medical management of genetic disease, and moreover the automation of the variant interpretation work for all decreases turnaround times. Specifically by patient phenotyping as well as literature curation. And with that I want to say that we hope that the most acutely ill newborns and children have access to this test as routine standard of care as that makes the difference.
Thank you to the RCIGM team as well as to all our collaborators to make it happen. Thank you.