Automate, Import, & Interpret: Using Custom Scripts in VSClinical

Editor's Notes

• #1: Casey’s intro
• #3: Casey’s intro
• #4: To get started today, I want to first express our appreciation for our grant funding from the NIH. The research and development efforts for a number of our software capabilities have been supported by the National institute of general medical sciences of the national institutes of health under the listed awards, as well as local grant funding from the state of Montana. Our PI is Dr. Andreas Scherer who is also the CEO at Golden Helix. I must mention here that the content described today is the responsibility of the authors and does not officially represent the views of the NIH.
• #5: Before diving into the topic of today’s webcast I would like to take a moment to give our attendees, especially those that may be new to Golden Helix, a brief introduction to our company. Golden Helix is a bioinformatics software company based out of Bozeman, Montana that has been serving customers all over globe for over 25 years. We began by providing research focused software for array-based analysis but early on shifted our focus to next generation sequencing applications and we have emerged as market leaders in the NGS space and now our focus is to provide high quality bioinformatic software that specifically enables our customers to conduct routine clinical applications for their NGS analyses. Our tertiary software solutions are scalable for routine work with gene panels all the way to whole genomes and are highly automatable to facilitate high throughput operations where large numbers of samples are being processed. This combined with our subscription-based business model, allows users to freely process an unlimited number of samples as needed, without the concerns about scaling costs that would be experienced with most per sample applications on the market. The assays designed in our software are flexible and user refined, and span a wide spectrum of applications, including somatic workflows for oncology-based analyses, germline workflows for hereditary cancer, inherited and rare diseases, prenatal testing, carrier screening analysis, family-based analysis, and pharmacogenomics analysis. Taking advantage of these capabilities is our very wide-spread global customer base – our users span government and testing labs, hospitals, universities, and many research and pharmaceutical labs. Our communication with our customers informs our software development as we aim to stay abreast of the most important features to develop, edge cases and different data types that we can support, and through this partnership our software has been regularly cited in reputable scientific journals, which is a testament to the work of our customer base.
• #6: Now that we have discussed some uses cases for our software, let’s review where our tools fit into the bigger picture of an NGS Workflow. The NGS workflow is divided into 3 stages where primary analysis encompasses everything from sample collection to sequencing, secondary analysis describes the processes for read alignment and variant calling, and the tertiary stage is where variant evaluation and reporting take place. VarSeq is a tertiary analysis tool that is designed to be agnostic to upstream sequencing platforms and secondary analysis pipelines, which means we accept NGS variant calling and alignment files from the various platforms and pipelines that are commonly used, provided that these adhere to standard VCF and alignment file formats. These upstream pipelines include Illumina and ThermoFisher, some of the emerging sources like MGI, and we accommodate PacBio and Oxford Nanopore long-read technologies. In fact, we are endorsed by PacBio as a tertiary analysis provider. We also have a long-standing partnership with Sentieon to provide labs with a secondary analysis solution as needed. VarSeq is one of the few platforms that can handle the range of variant calling outputs from upstream analysis pipelines, tackling both short and long read data, and scaling from small targeted panels all the way to complete whole genomes to accommodate the large number of variants analyzed and the increase in computational and storage demands. The graphical user interface of VarSeq serves as the front end for annotation and filtering, as well as clinical interpretation and reporting for variants, CNVs and fusions. However, we couple this GUI with our command-line-interface workflow automation tool - VSPipeline - for higher throughput processing for each component of tertiary analysis. Lastly, we provide robust data warehousing solutions via VSWarehouse which serves as a repository for aggregating and storing variant frequency data from your own cohorts over time. Warehouse facilitates efficient data management and enables easy retrieval of variant assessments or interpretations that can be applied to a growing cohort of samples and is deployed locally in your environment to enable data security. To learn more about automation and warehousing with Golden Helix, or any of the components of our software, we encourage you to review the collection of webcasts hosted on our site.
• #7: Each of the applications we just discussed has been diligently developed by our team here at Golden Helix, and we adhere to a highly structured and thoroughly documented manufacturing process. As a result of this commitment to quality Golden Helix is now an ISO 13485 certified medical device manufacturer as of January 2024, and is a CE marked medical device under IVDR as of April 2024 . This certification holds significant value for laboratories seeking their own ISO certification and IVDR compliance, especially those within the EU or those processing European samples, as the software can more easily be incorporated into a lab’s quality management systems and processes. Our certification and continued adherence to a robust QMS assures reproducibility of our quality specifications and manuals thus simplifying the validation process for any lab using current and future versions of our software. It is important to note that VarSeq is not CE marked for users by default – if a user desires to use VarSeq as a CE-marked medical device we have developed VarSeq Dx Mode which is available in VarSeq 2.6.1 and all future versions. When implementing this feature, we have a certification process which users must complete, and our support staff is ready to guide you accordingly through our user onboarding, installation and verification process and proficiency certification process as we move through the workflow validation process together.
• #8: As NGS analysis grows and labs scale from gene panels to whole genomes, there is increasing complexity in NGS testing deployment. Some of the topics that are interwoven into that complexity include increasing data volume, multiple data types (such as STRs and Methylation), and streamlining the reporting pipeline for complex variant interpretations. Fundamentally our solution is meant to meet these expectations and tackle this complexity when setting up your bioinformatic pipeline. The purpose of this webcast is to tackle one of these aspects of the complexity, and that is, tackling the customization of your bioinformatic workflow to make variant analysis your own. The point of talking about this challenge is to address high level constraints and issues that you have – if you have challenges in these areas, 2025 is the year that we at Golden Helix are tackling it. Stay tuned to all our webcasts this year so you can stay on top of all our developments and new features. _______________________________________________________________ The Challenge – Scaling Genomic Data Workflows Growing volumes of genomic and clinical data Increasing complexity of variant types (SNVs, CNVs, fusions) Need for secure, collaborative interpretation environments Challenges with file-based storage approaches Evolution from targeted analysis to comprehensive genomic workflows More samples, more mutation types, and growing metadata requirements Clinical interpretations must be reusable, traceable, and secure Legacy systems (file-based workflows, TSVs) don't scale Visual: Figure with simple icons representing security, collaboration, and storage challenges for this slide
• #9: More specifically, some of the considerations a lab may have about deploying NGS analysis include: 1. The installation – I. Historically, VarSeq was downloaded and utilized on a local desktop or server. This is an approach we still carter to. II. With our upcoming VarSeq releases, we now cater to several new development strategies – For one we have the ability to set up our secondary and tertiary analysis tools in the Warehouse workspace. This could be in a cloud environment, a server environment, or a hybrid application. For more information about this, our past few webcasts were on this topic. 2. The second stage for consideration is the degree of workflow Automation needed for a lab. III. Busy labs may require automation of the secondary analysis for FASTQ to VCF conversion. With tools like Sentieon, we have this capacity for you to do this through a traditional installation or through VSWarehouse. IV. Some labs may require the entire FASTQ to final report workflow automated as well. What this means is utilizing a broader scope of command line tools like cronjobs, triggered policies, SLURM processes – so that once a FASTQ hits a directory the VCF to project creation process runs). As our users aim to automate the tertiary portion of the timeline, they are likely going to be utilizing custom Evaluation scripts in VSClinical, which is what we are going to talk about today. Overall, we understand that one of the tradeoffs of a local deployment is a hands-on implementation effort. We facilitate this by dockerizing the software – however, the users still have the bioinformatic freedom to do what they want. For example, you are transitioning from array to VCF – this is an edge case scenario in which you need bioinformatic capability to converge all your workflows into one pipeline. The purpose of the webcast is to go into some detail and address some of the consistently discussed areas of customization and expose the audience to some custom capabilities and open your mind to some of these capabilities. We want to enable our users to be independent in customizing their bioinformatic pipelines. The focus of today’s examples will be in the context of the VarSeq GUI, using evaluation scripts to facilitate variant import, evaluation, and reporting.
• #10: So now, regardless of the deployment approach you take, let’s take a look at how the VarSeq software gives you a streamlined way to conduct tertiary analyses as efficiently as possible. The VarSeq software provides the interface that can create comprehensive workflows on both short and long read data, and our partnership with Sentieon also has you covered on the secondary analysis front if needed. ​ ​ It’s simplest to separate VarSeq into three steps. Step 1 is to import the full list of variants, SNPs and indels, and CNVs and SVs, from both long and short read pipelines. Imported variants can then be annotated against several data resources that are available within the software to build a variant filtering strategy. In this step we also enable you to see variant phasing and look at other features such as methylated regions. The goal is then by the end of step 1, there are a select few that are carried into the VSClinical interpretation hub in Step 2.​ ​ In Step 2, VSClinical houses the automated ACMG and AMP guidelines for evaluation of your germline and somatic variants. Here you will assess every available layer of evidence for a variant, draft and catalog comprehensive interpretations. Here are where our evaluation scripts come into play, tackling everything from importing data to creating an automated interpretations. Ultimately, we will complete step 3 which is to render your customized clinical reports. We ship a number of example report templates but know that users have a wide spectrum of report customization options that are rendered with a click of a button.
• #11: Backing out to a higher level view, we know that big data analysis can bring unique challenges. We don’t necessarily expect viewers to have run into all of these problems, these are just some common issues that come up when designing a workflow and that is what we are addressing. Some of the pain points we often see are: Importing variant records in bulk. Users may need to Sync multiple table sources or import legacy variant assessments which may be formatted differently than the VSClinical assessment catalogs. Importing atypical data files. Examples: Text files with fusions, unsupported data types, or external variant interpretations. Another common area is when users need to Customize variant interpretations for preferred fields instead of using the default VSClinical variant interpretations. Another large pain point is the area of Automated reporting. Customers report that with other pipelines, the route to report can be tedious and time consuming. They may also be limited in reporting formats, or they are unable to push their reports to other systems such electronic health record systems. With that, I would like to hand things over to Dr. Rana Smalling who will take you through how evaluation scripts can be used to address these and other pain points.
• #12: The purpose of this webcast is to show how our evaluation scripts can address some of the consistently discussed areas of customization, some of these pain points that Jennifer referred to, and expose the audience to some custom capabilities. We initially introduced eval scripts to import comprehensive genomic profiling data ( for example TSO500, Archer fusions, Oncomine and Ion Torrent Signatures) but the functionality is so versatile it can be expanded on for multiple other use cases, so we did just that. Now our scripts enable users to independently customize their clinical reporting pipeline. Evaluation scripts can help users save time, effort, and sanity, with automation for a wide range of imports. These imports are not limited to the CGP data, but can be expanded to importing your legacy assessments or variant classifications for example. They can be used to extend the standard capabilities of VSClinical, for example to automate the report section that a variant goes into based on a flag set from the variant table, they can be used to automate matching of drugs to clinical trials on a global scale. They can also be used to bring in non standard fields or to modify existing fields, or the behavior of certain processes in VSClinical, for example you can now use a script to render a custom interpretation which overrides the default auto-interpretation of VSClinical . Essentially, We have built this functionality for users to basically bring in any externally sourced clinical data into a custom report – the infrastructure is available for users to get very creative in designing their optimal clinical evaluation and reporting workflow.
• #13: So the next question might be, where can I use these scripts and how do I access them? The evaluation scripts are specific to the VSClinical stage (stage 2 that we mentioned earlier), which comes AFTER the VarSeq filtering and annotation stage. The script itself is a program, written in the javascript language, that automates process or tasks in the context of a VSClinical evaluation, so the scripts are based on fundamental functions within VSClinical. They are useful when we go from filtering to clinically evaluating impactful variants in VSClinical. At this stage, there are two aspects to consider- You want to automate as much as possible, but you don’t want let go of the user’s input and flexibility, so we give the user a way to automate in a way that addresses their specific needs for customization (We will highlight this on the demo going forward ------((from variant table to VSClinical) With these scripts, a user can maintain the fundamental functions in VSClinical, while implementing other aspects of their workflow such as importing IonTorrent or Archer data or data from other CGP kits, or customizing their interpretation.
• #14: So how does one go about customizing the scripts? We are not going to elaborate on scripting for the sake of this demo but two major points to make here are that: We have an existing library of scripts that are available to help the user get started (For example, we have scripts importing externally called variants like fusions, or genomic signatures like MSI and HRD) The scripts are fairly easy to modify for users with some experience with scripting. As mentioned they are programs written in the javascript language, so you can easily create a custom script starting with one of our existing scripts as a template. When modifying the scripts, a text editor like VSCode pictured here is very useful, because it adds some level of predictive assistance that helps you avoid syntax errors as it enables some awareness of the API type definitions that feed into our evaluation scripts. Of course, our support team is here to assist with this process as needed. The scripts are also stable and private, in that once you design a custom script, it’s yours to keep and reuse, similar to a report template. In this way, we give users the power to streamline their analysis and retain the flexibility to make it their own. At a high level, we’ve multiple users leveraging this valuable tool, and customers who have taken advantage of the feature enjoy the ability to customize their workflow.
• #15: The focus of today’s examples will be in the context of the VS GUI facilitating variant import, custom interpretation and populating assessment catalogs. We want to show, within the scope of a typical VS analysis where do these fit an and help facilitate the process: We will start off with the classic eval script – the import of TSO-500 genomic signatures and structural variants, then go into custom import of assessments and a custom variant interpretation. So Jenn is going to kick us off with an example using TSO500 comprehensive genomic profiling: *don’t go into too much detail in the script – show this is what it would look like if you did not have the script: before and after kind of, and show where the script is but do not open up and show 9you can say the script is accounting for type fields that you have to pull out of back end files and functions that you can use from our back end but if you contact us
• #16: VarSeq Client Perspective (why better) Most simple way, analysis Automating table exports to catalogs with Saved Exports  Defining a sample manifest using a catalog  Managing a blacklist catalog: Adding a variant and updating filters  Visualizing catalogs alongside variants, pileups and other VSWarehouse cohort frequencies  Converting a text-based custom database to a catalog  Tracking sample analysis progress by updating status in a catalog  Server-Side Perspective   Creating a new draft catalog  Assigning a draft to a reviewer  Approving and publishing as a reviewer  Customizing catalogs: adding fields, adjusting settings  Exporting a catalog to a file  Automated workflow example: Generating a report via REST APIs 
• #17: VarSeq is a tertiary analysis powerhouse for the VCF to clinical reporting pipeline for a variety of NGS testing use cases. Germline analysis: single sample, duos, trios, and cohort analysis. Somatic analysis: gene panels, somatic cancer kits, tumor/normal analysis. The VarSeq reporting pipeline can be augmented with clinical evaluation scripts to perform repetitive or novel functions. Classic use case: Importing fusions and genomic signatures. Current use case: Importing legacy data, creating custom variant interpretations, and more! There are near endless applications! Overall, We are facilitating your efforts by showing you these examples. That being said, please reach out to our FAS team if you are looking to commission us to build custom evaluation scripts for your lab.
• #18: We would once again like to mention our appreciation for our grant funding from NIH. The research reported in this publication was supported by the National institute of general medical sciences of the national institutes of health under the listed awards. We are also grateful to have received local grant funding from the state of Montana. Our PI is Dr. Andreas Scherer who is also the CEO at Golden Helix and the content described today is the responsibility of the authors and does not officially represent the views of the NIH. I would now like to hand the reins back to Casey!