The below text is a transcript from the webinar. Because it is a transcript, there may be oddities that arise from the process of translating speech into text. We recommend accessing the recording, above, to gain full context.
Overview
So with that, let me go ahead and get started. This first slide, as you read it, is a quote from one of our founders at Washington University, John Pfeifer. The essence of it is that molecular diagnostic testing is currently fragmented and we need unified solutions as well as a single end to end workflow that'll really enable us to scale and provide the promise of precision medicine. So, our mantra within Pierian is that we provide a one space solution that has no boundaries that allows you to provide molecular diagnostic testing and facilitate that testing across diseases.
PierianDx actually spun out of Washington University in St. Louis. A group of us, including myself and John Pfeifer within the departments of pathology and genetics, created a clinical genomics laboratory back in 2011 called Genomics and Pathology Services. Within that unit we launched what we call the Clinical Genomics Workspace, or CGW, under my leadership and within my laboratory or the CBW, and that was the initial product that we spun out of Washington University and formed PierianDx in May of 2014.
Since May of 2014 to present we've made multiple releases of CGW, added professional service offering s, have got series A funding as you can see on the slide, added partners to our PierianDx network north of 50 customers now, acquired a company called Tute Genomics that I'll talk a little bit more about and as you can see the latest has been to now add clinical microarray to the CGW offering.
We have a healthy mix of academic medical centers, cancer centers, commercial laboratories, reference laboratories, pediatric hospitals, and so we're very excited to be a part of this network and part of this partnership where a variety of different segments of the market trust us to help facilitate their molecular diagnostic testing.
Challenges in Molecular Diagnostics
I want to start by talking about workflow barriers in the molecular diagnostic world. As we mentioned earlier the first is a very fragmented workflow that many of you that have worked through-- the multiple screens open trying to work through a case all the way from accessioning a sample to signing out a report.
The second major barriers that you know...there's a rich knowledgbase that's created as you review and using out cases that could be re-used but that is frequently not available in laboratories because all you've got is a laboratory information management system (LIMS) that is generating reports.
The third is that even if you do have a very mature knowledgebase, in your own laboratory, you still are operating in an isolated environment where you can't leverage the combined community of medical directors who perhaps have seen a variant before and knowledge that you can then leverage because of the work that they put in when they review the variant. And finally, even with that sharing of a knowledgebase, there's still a laborious review of variant through a multiple databases including the literature for a pathogenicity or a clinical action ability assessment.
We really address those variants so the first is that we're integrated, we're streamlined to go end to end, enabling you to launch a molecular diagnostic in your laboratory very quickly. We have product that I'll describe that allow you to bring testing in-house much more rapidly and scalably than you could perhaps on your own. I'll talk a lot more about our knowledgebase it's constantly updated based on the newest available information. And finally I'll also talk about the collaborative component of our system in our network where interpretations and other components of are shared.
The Clinical Genomics Workspace
Today I'll focus on the CGW as well as in the end on the interpretation services. CGW is an agnostic software system supporting multiple instruments, thermal fissure instruments, assays for both of those vendors, somatic cancer and constitutional assays on the NGS side and the CMA side and it is broadly applicable for amplicon-based assays, hybridcapture based assays, things of that nature.
Our acquisition of Tute Genomics about a year ago now, greatly enhanced our constitutional capability including a tool call Phenolyzer that enables scoring of phenotypes and variants based on phenotypes that are entered in a patient's case.
Our proprietary knowledgebase is continually curated and aggregated and it is organized into multiple sources that are cleanly versioned, and version tracked and they include both public and proprietary sources. The partner sharing network is sharing of multiple components. One is really knowledge about assay design and assay validation, and secondary analysis of assays are all shared so that common partners wanting the same test can utilize best practices with respect to secondary analysis pipelines.
Clinical interpretations are also shared so these are plasticization as well as interpretive text associated to one or more variants in the context of the disease, and those then can automatically infer as you review your own case. So a variant that's been seen at another site and has been published is then visible in your report for consideration.
And then finally de-identified aggregated patient data are shared so that you can view variant frequencies by tumor pipes, see how often a variant was classified as actionable or pathogenic versus benign. Things of that nature.
So at the very base of our system are publicly available sources so multiple human genome builds, build 37, build 38, multiple genome protein models that you can pick and chose from as you decide how you want to report out on an assay.
Minor allele frequency databases are all available. And you can pick and chose again which of those and what cut off levels you want to use for...for example calling common variants. The next level is what we call our cleaned publicly available sources so these are databases like COSMIC, TCGA or ClinVar, Emory's database and Invitae's database as well as a variety of others that we normalized to disease encoding.
Highly curated proprietary database include NCCN and ASCO guidelines, FDA approved labels, clinical trials, that update on a weekly basis as new trials are released or are no longer recruiting as new NCCN flash updates are presented or as new updated drug labels are available. In addition, we have automated scoring of variance in the constitutional setting using the ACMG guidelines. That is automatically scoring on a subset of the evidence codes, allowing you to complete the scoring based on manual review of the rest of the evidence codes and then final calculation that's automated to assess the pathogenicity of a variant.
Finally, the proprietary data I've already talked about but essentially, again, shared clinical interpretations as well as de-identified aggregated patient variant frequency data that's associated patient demographic and clinical information such as disease. What I'm going to now do is really drive and show you much of the CGW features using the clinical micro array as the example setting. I'll highlight where there are common workflows and components with NGS. The workflow integration allows you to submit processed micro array results from a batch of samples. I'll show you how that's done in slides that are upcoming. This is very similar to the NGS workflow where we integrate very deeply with sites through site side VPN tunnels and directly to sequence servers, so we can access results of sequencing directly. The second features really bend the ability to review CNV and LOH segments and associated quality information using a sortable and filtering interface and this is very similar to how we do work in the NGS world as well. i.e., evaluate a variant to determine whether it's "real" or not.
The third is to then set laboratory specific validation filters. Again, this is common between CMA and NGS so that based on your validation criteria, you can view variants or segments that have passed validation. The fourth is really to view data in a genome browser based view and so that allows you to view quality data such as log2 ratios, allele difference plots. On the NGS side, this would be done using things like IDV to look at reads, presence of alternate Alleles within reads, a soft clip reads for fusions for example, things of that nature. The other feature in the genome browser is to then view annotations from databases like DGV and ClinVar so that's in the CMA setting. On the NGS side, you can view this information from, as I mentioned, ClinVar and DB Snip and Exac and ESP, COSMIC, TSGA.
Then finally, you can view segments or variants that are interpreted in prior cases and that's not only from your own organization but also across the PDX network. I'll show you some screenshot examples of that. Just like in the NGS setting, you can create clinical interpretations automatically and then that can automatically infer in future cases, for review and editing. A new feature that's now available in the NGS, and the CMA setting is to create query and review the published literature, and I'll go through a lot more detail in that where you can search for things like cyto genetic positions, genes, phenotypes, amino acid syntaxes, diseases, and drugs. Finally, both for the CMA and the NGS workflow, you can edit and finalize a report using a variety of templates and with both of these features obviously, we have our HL7 integration at the front end, and the backend to receive an order and sign out the order and pass that then signed out report to the LIS or the medical record, automatically.
Application Programming Interface (API)
We also have our API integration that allows you to do the same and then receives signed out order content for both the NGS and the CMA workflow. In the case of CMA, using either Affy or Agilent software. Those are the two micro array products that we support today. In the future, we're looking to support additional vendors. In using one of those two software solutions, however, for the arrays that you're running, you would follow your standard laboratory operating procedure through standing and segmentation analysis. You can then export the results from the vendor tools and then upload a batch of results via a run uploader tool. This is the same tool that's used in the NGS setting to upload primary sequencing data as well. CGW will then automatically analyze that batch and generate draft reports so that you can review that information in the different user interfaces that I described previously.
NGS and Chromosomal Microarrays
The differentiators here again, were highlighting what can be done for CMA but all of these really apply to the NGS setting as well. That is again, MRLIS integration via the HL7 or the API approach. A seamless workflow that moves from the vendor based software solutions to our solution that goes all the way to sending the data to us to report the data without user intervention, report finalization again, within our system, so that you don't have to go into any other systems as you review the case and edit it and sign it out. Again, those signed out reports can be accessed via the HL7 and API can then result those reports to your medical record.
A new feature then again, was to allow you to rapidly and robustly query greater than 18 million publications using, in a CMA setting, cyto genetic positions, genes, and phenotypes. In the NGS setting as well using gene syntaxes, diseases and drugs. For CMA as well as NGS, if you've got a legacy interpretation database, we could load that database into the system as you start so that you could immediately start leveraging those interpretations as you start to sign out CMA cases perspectively.
Then finally, as I said earlier interpretation services are available and, I'll describe what they are at the end of the presentation, but essentially varying scientists, all the way to medical directors sign out services that we provide with within PierianDx, where the best mix of technology and human expertise are used to deliver the near final comparable reports to you.
These are the interfaces that are used to review CMA cases and again the same interfaces are available for NGS, and you can look at, copy them or alterations. You can sort and filter by, things like the number of probes that have evidence of the CNA, the size of the CNA, the number of genes that are overlapping that alteration, genes that have been identified by OMIM that are overlapping, thing of that nature.
Again in the NGS setting you can do the same, with appropriate data and annotation criteria there, things like depth, zygosity, minor allele frequencies in different databases, things of that nature.
You can also create and save filters based on these criteria, so that those filters show up in all future cases of that same panel, or you can save case specific filters.
This example shows you the genome browser that's associated with CMA and here what you can see is individual segments here, in this case the segment that's been called out, the weighted log two ratio, where you can see the actual loss of a copy down here, you can see the allele difference here, where there's compression of the two alleles, to homozygous allele that are left, versus the three pattern that you can see on either side.
In addition, you can then see annotations, that are genes that overlap as well OMIM genes that are overlapped with the segment, as well as data from other knowledge bases such as Clinvar DECIPHER syndromes.
There are additional tracks in the same genome browser view, where now we're showing legacy interpretations that have been loaded, and so there's a my legacy cases track that lets you look at segments that you've seen before and that we've preloaded so that as you start using CMA within our system, you haven't lost any of the prior knowledge that you had in whatever workflow you used before. And then finally this track that shows all institutions interpretations, shows you segments that have been interpreted before by anyone that is using the system within our network and is opted into sharing those interpretations.
I want to spend a few minutes now, talking about the publications' knowledge base that I've been referring to. So this is a new feature that was released in 6.1 and both the use cases, and the features are really summarized here. So the first is that you can find publications of interest based on six categories, gene, amino acid syntax, disease, drug, phenotype and cytoband. You can also do free text queries. These categories, however, have been created so that you can very robustly search the published literature based on gene aliases, diseases that use oncology subsumption drugs, and phenotypes that also use oncology subsumption, in order to find papers much more robustly and accurately then you could with PubMed alone.
The second thing that you're able to do is filter by a variety of things such as publication date, amino acid ranges, so even if you can't find papers that mention your specific amino acid chains that you have in your patient case, you can look for papers that have variants that are near the variant that are in your patient case, and you can also filter by additional facets. Things like clinical relevance, look for papers that have part of FDA approved labels or guidelines, study types, so papers that are the result of meta-analysis, or late-stage trials versus case reports. Papers that really are filtered by variant types, are they talking about indels or single leukocytes variants. You could find papers that are mention amino acid syntaxes, which are unframed fusions versus framed sifts, things of that nature, by conscious, and finally by inherence pattern, looking for paper that are talking about a particular gene as well as that gene being mentioned in that paper, working in an auto zonal recessive manner, for example.
You can also, obviously, view publication content, so titles, abstracts, full text where available. You can also link out to PubMed and the full text, and that's as available to you by licensing agreement, at your organization. In addition, you can rank categories based on the number of articles that are co-mentioning two or more terms. So for example, you can the top disease based on doing a search on a particular gene, so you can use this as a discovery interface as well, to say hey, which diseases or phenotypes are top ranking based on this gene, and you can do that with any of the categories that are available, and I'll show you examples of that.
Somatic and Germline Workflows
This supports both germline and somatic workflows in NGS and CMA. Now I've kind of given examples here, so if you're looking at a somatic cancer essay, you can do a disease or phenotype search on the tumor type. A gene and an amino acid syntax range search, look at the top ranking drugs, read those papers that meet those requirements and then use that to build your interpretive text. In hereditary cancer setting, you can also do those types of same searches, and now looking for risk profiles so that this is now risk of cancer, not somatic cancer.
In the germline setting, you can do germline disease, gene, and amino acid syntax to range in the sort of more broad excel based essay, or large panel based essay, that are more phenotype based. You can do gene and/or phenotype based searches, including amino acid syntaxes and range, or not.
Finally, in the CMA setting, you can do cytoband and phenotype searches, or gene and phenotype searches. I think almost all of these examples are shown in slides that follow here.
In the CMA setting, where we're looking for papers that are mentioning a particular cytoband, and that cytoband overlapped your variant of interest, or your segment of interest, and you can then rank the top ranking genes, as well as the top ranking phenotypes that I'm showing there in the inset. You can then very quickly drill into phenotypes that have been mentioned as part of your patient's case, to look for evidence in the literature, that that cytoband is associated with that genome type.
This next example shows you an example where you can filter based on publication year and other clinical and variant facets, and as you can see in the bottom, then you can also link out to PubMed and full text using these hyperlinks down below.
In the somatic cancer setting you can obviously, as this example shows, search on a gene and amino acid syntax, you can then see the top ranking drugs, top ranking genes, you can filter by, in this case, I've done a guideline only clinical relevance filter, so you can look at the paper.
This case I've done a guideline only clinical relevance filter so you can look at the papers that have been mentioned and guidelines, and so then you can review those papers as you write your interpretative text.
In this hereditary cancer example, again I've done a gene and syntax search, but now by clinical relevance of risk and I can then highlight the top ranking diseases and again I can read about the papers that are talking about risk of that cancer in this example.
In the XLN based work flow, I can do a gene and amino acid syntax search, and again, look at the top ranking diseases or phenotypes, and then I can review the references to the amino acid changes in the manuscript in order to them write my interpretative text.
And then this final example then, shows you that same thing where we're now just doing a gene search and doing filtering across amino acid positions and then again, ranking genes, looking at particular syntaxes that may be close to the syntax of my patient's case in order to then write the interpretative text.
Report Writing
Now during report editing and writing, as you know, you can save a clinical interpretation, and a screenshot of that is shown here, again in the CMA setting, so in this case using the ISCM nomenclature, the variant was automatically identified and you can write what's called a coordinate based rule in order to then have a classification and an interpretation infer automatically. In this case we're writing a coordinate based rule on chromosome 11, where a start and stop coordinates are the coordinates that have been identified in that particular patient segment. And where we're requiring a copy number of between zero and one, so at least a loss of one copy, and then we're classifying that as pathogenic with the appropriate interpretative text I've shown down below.
You can also write as you may know, variant based rules and so those variant based rules then are available in the protein syntax as well as other syntaxes, and that then can be ... inferred in future cases.
This just summarizes and again, in the CMA setting, an example of a report template where you can edit all components including the actual nomenclature, the result, the automatically calculated gender based on the data. And templates like this are available on the NGS side, as well as those that look radically different that show variant summaries, and result summaries of variants in the somatic cancer setting, for example, that allow you to discretely delineate those variants that are responsive to therapies or have prognostic implications to different cancers, or have diagnostic, or have a risk of that cancer as well as trials that are relevant. So again, take home message here is that you're able to edit your report using a variety of report templates. Some that may be very text based, others that may be very graphically and table based.
Interpretation Services
I want to close by talking about our Interpretation Services. So the interpretation services really are composed of expert teams of somatic directors and what we call variant scientists who leverage the CGW technology in order to review a customer patient cases. And this can be added on, and I'll show you kind of the differences between CGW and you doing interpretation on your own, or add on interpretation services that we offer. And again, it's a very rapid turn around time. And it can also complement your existing resources and this can be due to expertise or peak demand times where we can come in and help, or we can continually help through this process in order to decrease your turn around time.
It's the mix of technology and human medical expertise that we believe is necessary for a high complexity test like NGS and CMA where with CGW in the NGS setting, you obviously get automated classification of variants using our knowledge base that I've described. In the cancer setting getting suggested therapeutic options and risk information in the hereditary cancer setting. In using the PDX knowledge base in the constitutional setting then evaluating a subset of the ACGM evidence codes using our knowledge base. And then in the somatic cancer setting, doing clinical trial matching based on molecular findings, as well as the tumor type.
The variant interpretation services then that we would provide would then assess existing variant classification and interpretation and will update that based on that the newest literature. Again, we'll use our own publication feature here in order to do that. We will draft clinical interpretations for novel interpretations that ... I'm sorry, for novel variants that don't yet have an interpretation in our knowledge base. And I'll describe later why that can occur, as well as reclassified variants. We can evaluate pathways that are operating, especially in somatic cancer, in order to create more cohesive interpretations across variants. We can refine the trial matching based on pathway crosstalk and use a client specific SOP. So for example, you may want to rank trials differently at your site, i.e., trials that are offered at your site versus others, or trials by distance, things of that nature.
And finally, we can fully then refine the variant classification using established guidelines and evidence codes. In other words, we can do that work for constitutional cases using the ACMG semi-automated calculations and classifications that we have.
In the CMA setting, much of the same applies, so obviously you yourself can go through quality review through our genome browser that I described, as well as use the knowledge base to automatically classify variants based on your legacy interpretations. Interpretations you've written in the system, or interpretations that are available in the system throughout the entire network. And you can then evaluate variants using a variety of knowledge bases, as well use our publications knowledge base in order to draft or update interpretations for segments that are either novel or where knowledge has changed respectively.
With interpretation services, we can again do a variety of that work including the quality review, the clinical review of the segment, including assessment, using the annotations and the publications knowledge base, and based on AGMC guidelines, for chromosomal micro-array. And again, we can provide then the variant interpretations for novel segments that have not been encountered by the PDX. Network before, as well as case recommendations based on the assessment that we do in order for you to then finalize and sign at the report, or our Medical Director can actually do those last steps both in the NGS and CMA setting.
Future Enhancements to CGW
What we'd really like to do is again make the process of reviewing cases both in the NGS and the CMA setting much more rapid and much more facilitated, and even greater than what we're doing today. And the way we're doing that first is to facilitate saving and sharing of interpretations so the current state is that we have thousands of interpretations that are shared across the PDX network of over 50 partners. So that already is providing great value.
However there are some barriers. One is that as I showed you, interpretations don't get saved automatically when a report is signed out. You've actually got to, before sign out or after sign out, as a Medical Director, or a member of the clinical team actually save an interpretation within our system. So there's a process to do that.
The second is that sharing is hampered because the current workflow as effort on the clinical team, there's a process to what's called publish a variant, and that process doesn't fall nicely in to the clinical workflow, and so even if you saved an interpretation many times, that interpretation then isn't shared in the network. And so in order to really evaluate this issue, we've took all clinical actionable variants in our entire system, that is all cases that have ever been reviewed and signed out in our system, and where variants have been classified as clinically actionable by Medical Directors who signed those cases out, we then random through the shared knowledge base to see how often the classified variants were ... I'm sorry, how often the variants were classified based on the shared knowledge base, and what we found was that the results were that the knowledge based wasn't optimally leveraged as it would have been if the entire set of prime interpretations were shared.
In other words, some reports didn't have interpretations that had even been created. Others, the interpretations had been created but not yet shared. And so we've been working through an approach that I have detailed in the slide. Mainly we work with a set of our existing Medical Directors, and discuss this barrier. And the consensus was to really share very quickly, and rapidly right up sign out. And so the features that we're working on in the last quarter of this year are to one, automatically save interpretations concurrent with report sign out. In other words, if a report has been signed out, obviously the interpretations within that report have been associated with a patient case, and therefore now merit being shared.
And the second is to share those ... Save interpretations without requiring the publish step. So that things that are outside the clinical workflow have been stripped out, allowing folks to share more easily and therefore also to leverage content that's already within our system, but isn't being optimally utilized.
The second is that even with that process, as I said earlier, variants get interpreted when one or more Medical Directors or our variant scientists, or our Medical Director reviews a variant clinically. And what that does is for rare variants or where variants that perhaps were last interpreted 12 months ago or 18 months ago, or two years ago where there's new evidence available, those variants then need to be either interpreted again or interpreted in the context of the newest clinical evidence. In order to really then reduce even that work for our partners where we're pre-curating with our annotation and interpretation team all variants that have either established clinical evidence, i.e., guideline or FDA approved based, or have strong immersion clinical evidence in the literature across 85 genes that are relevant for solid tumors, as well as heme malignancies. And so this supports the vast majority of low end assays that are typically in the less than 50 genes that both or solid tumor and heme.
So this then will be released at the end of this year as well. So that the end effect willl be that more and more variants will require little to no review by Medical Directors who are signing out cases in these small panel solid tumor and heme cancer settings. And a minority of variants will perhaps require a novel interpretation or perhaps a change because vary late changing clinical evidence is now available.
What we've talked about today is an end to end solution for molecular diagnostic laboratories who are doing NGS or CMA assays, both in the cancer and constitutional setting. Providing integration with existing clinical systems that laboratory sites including the EMR and the LIS, IHL7, or API integration, providing an out of the box solution for NGS for secondary analysis support, and we can do that ourselves, or we can do that with partners. So for example, true site tumor 170 assay, that's actually been done with the alumina app that analyzes true site tumor 170, so we have a variety of ways to really support secondary analysis for your laboratory.
Using our knowledge base then to review and sign out cases and that knowledge base includes curated information as well as shared interpretation knowledge base.
And then finally our interpretation services which can facilitate your final review through variant scientist, all the way to sign out, should you need that capability with our own Medical Directors.
And then I've concluded by telling you about how that knowledge base and those shared interpretations are getting better so that your turn around time, your review time are all going down.
