We’re back from ESHG 2026 in Gothenburg, and Booth #464 turned out to be one of the better vantage points we’ve had in years for taking the temperature of the field. Thank you to everyone who stopped by, stayed for the longer conversations or simply grabbed a t-shirt. It was definitely worth flying across an ocean for this event. Going in, we framed our conversations around three themes: clinical yield, automation, and scalability. We came out with those three confirmed and a fourth that we didn’t expect to dominate the hallway talk the way it did. More on that at the end.
Clinical yield: long-read crossed from promising to expected
The shift we noticed most this year was in how labs talked about long-read sequencing. A year or two ago it was a research curiosity for most clinical groups. At ESHG 2026 it was a planning assumption. The questions weren’t whether to adopt long-read, but how to fold it into an existing pipeline without running two parallel operations.
That matched what we came to show. VarSeq runs short-read data from Illumina, MGI and other short-read sequencing provider. Strategically, we have a long standing relationship with long-read sequencing providers such as PacBio and Oxford Nanopore. Our workflows are optimized workflows for long-read variant calling, structural variant detection, and tandem repeat analysis, the complex regions short-read platforms tend to leave unresolved. The labs we spoke with weren’t chasing more data; they were chasing the variants that traditional approaches quietly miss.
Evidence quality came up just as often. The consistent theme: an interpretation is only as defensible as the evidence behind it. That’s the logic behind monthly-curated annotations from ClinVar, gnomAD, and premium sources including CancerKB, OMIM, LOVD, and Genomenon, and behind the Genomenon Mastermind integration that puts cited evidence directly into the VSClinical ACMG workflow.
Automation: the conversation has matured past “does it save time”
A few years ago, automation pitches at conferences like this were about speed. This year the labs we talked to had moved past that. They assume automation saves time. What they’re scrutinizing now is whether automation preserves the defensibility of a result.
That’s a healthier conversation, and it’s the one we wanted to have. VSClinical’s guided ACMG/AMP workflows score evidence alongside each criterion, so a classification carries its rationale with it rather than living in a separate document. VSPipeline moves a sample from sequencer output to a structured assessment report without an analyst working through every step by hand. And VSWarehouse archives each assessment, so the next sample with the same variant inherits the prior call. The repeated point from medical directors at the booth: in a regulated environment, automation cannot mean a black box. We agree, and that traceability is the part we’re proudest of.
Scalability: the same software, met at every size
We spoke with everyone from first-panel startup labs to national genome program staff, and the encouraging part was how often the same product fit both. Or sample-based licensing lets a small lab onto an enterprise-grade platform while larger institutions move to usage-driven pricing, and deployment scales from high-end on-prem all the way to full cloud. Twenty-five-plus years serving labs in more than 50 countries under an ISO 13485-certified quality system is the foundation that makes that range credible rather than aspirational.
The unexpected fourth theme: what the AI news did to the security conversation
Here’s what we didn’t have on our agenda but ended up discussing at the booth more than almost anything else.
ESHG 2026 landed in the same window as a wave of AI news: most notably the public release of Anthropic’s Claude Fable 5 and the restricted Mythos-class models behind it. The headline for our field wasn’t the coding or analysis capability. It was the cybersecurity story. These models are remarkably good at finding software vulnerabilities; as a defensive tool the results are striking, with Mythos-class models autonomously surfacing hundreds of flaws in widely-used codebases that had survived years of manual review. Anthropic was candid enough about the dual-use risk to withhold the most capable version from general release and to route sensitive security queries on the public model to a more constrained one.
That news clearly rattled around in people’s heads on the way to Gothenburg, because the questions we got at the booth had a different edge than in prior years. The insight a lot of attendees seemed to arrive at independently: the same capability that helps a defender find a vulnerability faster lowers the cost for an attacker to find one too. For most software that’s abstract. For a clinical genomics lab it isn’t. The data at stake is patient genomic data, the most identifying and least revocable information a person has. You can reissue a credit card. You cannot reissue a genome.
So the conversation kept circling back to deployment architecture, and that suited us fine. VarSeq can run entirely on-premises and most importantly in certain high-end configurations fully air-gapped, with no path for patient data to leave a lab’s own infrastructure. In an air-gapped deployment there’s no external network surface for an AI-accelerated attacker to probe, because there’s no external connection at all. We’ve always treated data sovereignty as an architectural commitment rather than a feature line, and this year, for the first time, a lot of the people we spoke with were arriving at the same conclusion on their own. The industry’s default has been to push everything to the cloud and trust the guardrails. That’s a defensible bet for many applications, but for irreplaceable patient data in a world where vulnerability discovery is being automated, removing the external attack surface entirely is starting to look less like caution and more like common sense.
It was the most interesting strategic takeaway of the week.
ESHG 2026 was a strong conference: excellent questions, mature conversations, and a field that seems to be thinking hard about both clinical rigor and data security. Thanks to everyone who made the trip to Booth #464.
#ESHG2026 #Genomics #GoldenHelix #PrecisionMedicine #EnterpriseGenomics
