Automating Pharmacogenomic Workflows with VSWarehouse 3 From Variants to Clinical Reports

Editor's Notes

• #1: Happy to be here! I can't wait to show everyone our exciting new automation capabilities for PGx.
• #3: Casey’s intro
• #4: Before we dive in, I want to take a moment to discuss who we are. Golden Helix is a bioinformatics company founded in 1998 with over 25 years of experience in developing software to support both research and clinical genomics. We have a strong global presence that spans a large customer base including hospitals, universities, government organizations, pharmaceutical laboratories, and research institutions. Our mission is to deliver robust, accurate, clinical-grade genomic software that is capable of handling increasing sample volumes and supporting a broad range of use cases including oncology, hereditary and rare disease analysis, prenatal testing, carrier screening, and of course, pharmacogenomics. One of our key differentiators in the market is our support for local, on-premise deployments, along with the ability for users to have full control over their workflow design and execution.
• #5: Now let’s talk about where we fit into the overall NGS pipeline. Our VarSeq software suite serves as a tertiary analysis platform that handles the annotation, filtering, evaluation, and reporting of variants. These variant calls may come from any number of secondary pipelines such as GATK, DRAGEN, or our partner platform Sentieon. One of VarSeq’s strengths is its pipeline-agnostic design. It can accept variant data from any upstream workflow and supports a wide range of sequencing technologies, including both short-read and long-read platforms. The VarSeq suite is also highly scalable. Whether you’re analyzing small targeted gene panels or performing whole-genome analysis on somatic or germline data, VarSeq provides the tools and performance required for high-quality interpretation. And because genomic analysis can be computationally demanding, we offer the flexibility to deploy VarSeq and its associated pipelines in the environment that works best for you.
• #6: This is why we recently focused on expanding on our traditional on-premise deployment model to also support cloud-based deployments in any AWS or Azure environment of your choosing. Of course, on-premise and even fully air-gapped installations remain fully supported and will continue to be, ensuring that organizations can retain complete control over their data, infrastructure, and security policies. But for labs looking to scale beyond local hardware or staffing limitations, the cloud offers a powerful alternative. By leveraging cloud infrastructure, users can tap into virtually unlimited compute capacity, an especially important advantage as sequencing throughput increases and larger, more complex datasets like long-read genomes become more common. This expanded deployment flexibility allows organizations to choose the environment that best aligns with their operational needs, whether that’s maximizing scalability, preserving data sovereignty, or achieving a balance of both.
• #7: Finally, I want to highlight our recent ISO 13485:2016 certification, as of January last year, reflecting our commitment to a robust quality management system integrated with our manufacturing processes. VarSeq has been recognized as a CE-marked Medical Device under IVDR as of April 2024, ensuring required compliance for labs processing European samples. With the VarSeq Dx Mode now available, our certified support team is prepared to assist with user onboarding and workflow validation, facilitating a seamless integration for your laboratory needs.
• #8: Now that you know a little more about Golden Helix, let’s take a moment to define what we mean by Pharmacogenomics, or PGx. Pharmacogenomics is the intersection of pharmacology and genetics. It seeks to understand how an individual’s genetic makeup affects their response to medications. Genetic variation in key gene families, especially those involved in drug metabolism, can have a significant impact on how drugs are absorbed, distributed, metabolized, and excreted by the body. These differences can influence both the efficacy of a drug and the risk of side effects or toxicity. PGx testing offers several important clinical benefits. It can help: - Predict how a patient will respond to a particular drug - Guide drug selection and dosing decisions - And identify individuals who may be at risk for adverse drug reactions When interpreting PGx test results, there are three primary components to focus on: - First, the diplotype, which reflects the individual’s genetic results for a specific gene. These are typically expressed using star allele notation, such as *1/*4. - Second, the phenotype, which translates those genotypes into a predicted drug metabolism status—like poor, intermediate, or ultra-rapid metabolizer. - And finally, recommendations, which link that genetic and phenotypic information to specific clinical guidance—whether that’s a dosage adjustment, an alternative medication, or increased monitoring. Together, these components form the foundation of a personalized approach to drug therapy, helping clinicians make safer, more effective treatment decisions.
• #9: NGS is an excellent tool for pharmacogenomic testing because it can perform this analysis using existing exome and genome sequencing data. However, it should be noted that there are a number of intronic and intergenic variants required to call important star alleles that will be left out of an exome analysis. The first challenge in the utilization of NGS data for pharmacogenomic testing is the determination of which specific named alleles are present in each pharmacogene. These alleles are typically described using star allele notation, which identifies pharmacogenomic markers by means of a designated number for a given gene. For example, the star allele notation for the deletion of a single amino acid at coding position 775 in CYP2D6 is *3. Many star alleles represent a combination of several variants when compared to the human reference sequence and some are defined by structural variations, such as gene fusions and full gene deletions. For autosomal genes, allelic determination algorithms assign a diplotype for each gene, specifying a specific combination of two haplotypes or named PGx alleles. Consider the gene CYP2D6. This gene is a excellent example to discuss as it is highly polymorphic and involved in the metabolism of many commonly prescribed drugs. With over 170 star alleles defined there are thousands of possible diplotype combinations. However, only a few hundred diplotypes are associated with a non-normal metabolizer phenotype.
• #10: CYP2D6 is one of the most technically challenging pharmacogenes to analyze due to its high sequence similarity with the CYP2D7 pseudogene. This homology creates difficulties for traditional variant calling methods, particularly when it comes to distinguishing between the two genes. On top of that, accurate genotyping requires specialized techniques to identify complex variation, especially copy number variations and structural variants, which are common in CYP2D6 and can significantly affect drug metabolism. To address these challenges, we developed CypCall, a tool specifically designed to call CYP2D6 star allele diplotypes from whole-genome sequencing data. It works directly with aligned reads in BAM or CRAM format and uses algorithms tuned for these highly homologous regions. CypCall is capable of detecting the full spectrum of relevant genetic variation in CYP2D6, including SNVs, CNVs and SVs. The results from CypCall can be directly imported into VarSeq, and incorporated into VSPGx for the identification of relevant recommendations and inclusion in generated reports.
• #11: So how do we translate detected genotypes into actionable recommendations? That’s where VS-PGx comes in. VSPGx provides a set of tools within VarSeq to support pharmacogenomic workflows by simplifying the calling of named alleles, the annotations of relevant recommendations, and the reporting of pharmacogenomic findings. The two primary components of this application are the PGx Variant Detection and Recommendation algorithm and the PGx Report Generation system. The Variant Detection and Recommendation algorithm identifies pharmacogenomic diplotypes and annotates them against drug recommendations. This algorithm begins by identifying the named alleles present in the sample. For autosomal chromosomes, this process consists of identifying the best matched diplotypes, which consist of a pair of named alleles for each gene. Once diplotypes have been assigned for each gene, the algorithm matches these diplotypes to phenotypes and recommendations. After running this algorithm, a clinical report can be generated using VarSeq’s customizable reporting system. Clinical reports are generated using an easy-to-modify Microsoft Word report template and VarSeq comes shipped with an initial PGx report template that serves as an excellent starting point for creating custom reports. Information included in this report includes Implications for Current Patient Medications, Gene-Drug Interactions, Prescribing Recommendations, Phenotype Associations, and a description of all tested alleles. The calling of diplotypes, annotation of phenotypes, and reporting of recommendations is performed by VS-PGx in just a few simple steps with minimal user involvement. While the annotations and report templates can be customized, the annotation tracks and report templates provided by VarSeq have everything you need to start annotating and reporting all alleles defined in the CPIC database.
• #12: Our pharmacogenomic annotations in VarSeq are built on two sources: CPIC and the FDA. CPIC provides peer-reviewed clinical guidelines that link specific genetic variants to phenotypes and prescribing recommendations. These guidelines include standardized translations from genotype to predicted metabolic phenotype, as well as graded evidence levels that help guide clinical action. The FDA also plays an important role in PGx guidance. Many FDA-approved drug labels include recommendations based on the genotype of specific genes. To create a more complete and clinically actionable resource, we’ve augmented our CPIC annotations with manually curated FDA pharmacogenomic recommendations. This combined annotation set ensures that VSPGx provides robust coverage for a wide range of clinically relevant genes and drugs. But we also understand that no two labs are exactly alike. Some labs may follow local guidelines, include additional genes, or have different reporting preferences. That’s why all the annotation sources used by VSPGx are fully customizable. Labs can supplement the standard annotations with their own data—adding new genes, diplotypes, prescribing recommendations, or even modifying genotype-to-phenotype mappings—so they can generate PGx reports that align with their internal protocols and clinical goals.
• #13: VSWarehouse 3 is Golden Helix’s complete server platform for scalable, automated genomic analysis. It’s designed to centralize and streamline your informatics workflows, from alignment and variant calling to variant interpretation and report generation. One of the key strengths of VSWarehouse 3 is its flexible deployment model. You can deploy it in your own cloud environment, whether that's Amazon Web Services, Microsoft Azure, or your institution’s private cloud. We also support on-premise deployments, giving you full control over data privacy and infrastructure. VSWarehouse 3 is built for workflow automation. It supports automated execution of VSPipeline, Sentieon, and other bioinformatics tools, including your own custom scripts. The platform also integrates seamlessly with external cloud vendors and institutional data stores, enabling a smooth, end-to-end data analysis pipeline. It also supports the full range of Golden Helix analysis applications, including VarSeq and VSClinical along with any custom applications your team may need to run. This makes it a versatile hub for both research and clinical operations.
• #14: Let’s take a closer look at the workflow automation capabilities of VSWarehouse 3. The platform supports a fully automated, end-to-end analysis process, starting from raw FASTQ files all the way through to final reports. This minimizes the need for manual intervention and helps labs scale up without increasing staff burden. It also includes built-in pipeline tracking and version control, which is essential for both clinical and research settings. Every analysis step is logged, and pipeline versions are recorded, ensuring that your results are not only reproducible, but fully auditable; an important requirement for regulatory compliance and accreditation. VSWarehouse 3 is built on a cloud-scalable architecture, allowing it to handle large volumes of data and run multiple analyses in parallel as needed. Whether you're processing a handful of samples or thousands, the system can dynamically scale to meet your throughput demands. Finally, our use of containerized workflows with Docker ensures that each analysis runs in a consistent and controlled environment. This makes it easy to deploy and maintain workflows across different infrastructures: cloud or on-premise.
• #15: Now let’s break down the individual components of an automated PGx pipeline starting with sequencer output. The pipeline starts with the processing of raw sequencing data to generate FASTQs which serve as the basis for alignment and subsequent variant calling to generate a BAM and VCF files. This makes up the secondary analysis component which is typically run in a command line fashion using tools like Sentieon. The VCF and BAM serve as inputs for tertiary analysis, where a number of annotations and algorithms are leveraged for variant filtering. This analysis also includes CYP2D6 genotyping performed on the BAM files via CypCall. The filtered variants and CYP2D6 calls are then evaluated by the PGx annotation algorithm which performs genotyping and identifies associated phenotypes and recommendations. Once this is complete an initial report is generated. This entire process is completely automatable but requires some effort in initial setup and design to configure and validate the secondary pipeline and tertiary analysis workflows. Fortunately, VSWarehouse 3 has been designed to dramatically reduce initial implementation by providing users with pre-configured workflows which serve as an excellent starting point. Workflows can be customized and maintained directly in VSWarehouse, which serves as a backbone for centralizing the entire PGx pipeline, enabling the automation of the entire process from FASTQ to report.
• #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: Before we wrap things up and begin taking questions, I wanted mention our appreciation for our grant funding from the 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 Hayley!