Over the last month, Illumina, Pacific Biosciences, and 10x Genomics have all announced new software projects that will bring genomic data analysis and sharing to cloud computing, and the rise of AI is driving this shift in multiple ways.
In last week’s newsletter, I wrote about Illumina’s new BioInsight platform, a single entry point to its collection of digital products that also changes how customers pay for access.
Illumina Launches BioInsight, the Costco of Genomic Data: Newsletter, Oct. 14
Today, Illumina announced a new software product called the BioInsight Platform, which is not to be confused with BioInsight, the name of the business segment created a year ago that oversees its data analysis and data generation products and programs. Or maybe it is?
Today, as part of the company’s presence at this week’s 2026 meeting of the American Society of Human Genetics and Genomics in Montreal, Illumina announced early-access to three new capabilities within BioInsight.
DRAGEN Ultra aims to help researchers do more than just call variants. Using Illumina AI models, it can prioritize candidate variants and sort large numbers of research samples. It will use dnaAI, a new AI model developed by Illumina that predicts variants most likely to explain certain observed phenotypes, as well as the company’s other AI models, such as PrimateAI-3D and SpliceAI2.
“Applying dnaAI to large scale datasets resulted in an 8 percent increase in disease associations across unsolved cases,” Illumina said in a statement. “The model also ranked previously established disease-associated variants among its top four candidates in 99.4 percent of genomes analyzed.”
BioInsight Platform Learning Cohorts, which draws on large datasets available within the platform, can help researchers put together cohorts based on specific genomic features or phenotypes. And the BioInsight Workflow Orchestrator helps connect sample tracking, sequencer run planning, quality control, and secondary and tertiary analysis.
“I would expect a few years from now more of us go cloud-first than we do on-instrument or on-premise to overcome some of these different challenges,” PacBio CEO MarkVan Oene said.
Earlier this week, PacBio announced a partnership with Verily Health to build a workspace for the company and its customers to upload long-read sequencing data for analysis and sharing. I’ve already written about how it could lead to a slight shift in how PacBio charges for its sequencing products.
But it may also offer a way for researchers to increase the size of their studies without having to obtain and sequence samples themselves.
“We don’t have hundreds of thousands and millions of genomes out there for people to access, and so we’re trying to create the software infrastructure so that people can access and utilize and learn from our data more seamlessly,” PacBio CEO Mark Van Oene told Ion Genomics. Moreover, putting data in a place where AI model builders can access it and use it for training models could pay dividends down the road: if they’re trained on a particular type of data, the more likely users of those models will have to generate that kind of data to get the most out of those models, he said.
But AI is also pushing genomics companies to embrace cloud computing for a more fundamental reason: cost.
The building spree of datacenters to support AI throughout the US has led to higher costs of basic components like data storage and memory, not to mention graphics processing units. PacBio has seen those costs go up threefold over the last year, bringing on-premise computing up to where cloud computing costs are now.
“It’s making the cost of running those analyzes [on the cloud] much more attractive for us, so we’re pushing a lot of efforts into cloud-enabled sequencing technologies in the future,” Van Oene said.
High-throughput sequencers and spatial platforms like 10x’s Atera feature huge amounts of computing power onboard, just to condense everything into a datafile that users can run though their own bioinformatics pipelines.
“I would expect a few years from now more of us go cloud-first than we do on-instrument or on-premise to overcome some of these different challenges,” Van Oene said.
More news out of ASHG
Ultima Genomics has launched 150 bp paired-end sequencing on its UG200 sequencer and is releasing its Solaris 2.0 Max sequencing consumables, which double output per wafer to up to 20 billion reads.
“[These] are the two capabilities our customers have asked for most as they work to deploy Ultima’s sequencing technology across more applications, especially germline applications requiring clinical quality,” CEO and Cofounder Gilad Almogy said in a statement.
The changes enable a $60 human genome at 30X coverage and paired-end sequencing makes Ultima WGS data more similar to what researchers are used to analyzing (paired-ends aid with mapping and aligning reads.)
On Monday, Ultima announced it would provide sequencing readout for the single-cell aspect of a new 80,000 sample immune profiling study out of the UK Biobank.
UK Biobank 80K-Sample Immune Study Taps Ultima Sequencing, BD Single-Cell Platform
The UK Biobank is collaborating with IMU Biosciences, Ultima Genomics, and Waters on a multiomics immune profiling project that will analyze at least 80,000 samples.
This follows Ultima’s selection to support Precure, a joint venture led by Mayo Clinic and Thermo Fisher Scientific to create multiomic data from one million samples. Helix, an Illumina spinout, will generate whole-genome data on a fleet of UG200 instruments.
Complete Genomics is showcasing expanded sequencing capabilities of its CoolMPS 600 kit and DNBSEQ T7+ instrument. CoolMPS 600’s SE600 kit extends single-end reads to 600 bases with Q40 accuracy, up to 3.0 terabases per run, and independent dual-flow-cell operation.
In a company-sponsored CoLab session, Andrew Carroll of Google Research said pangenome mapping with DeepVariant models trained on CoolMPS 600 chemistry brought errors on the T2T v5 GIAB truth set for HG002 below 50,000, a more than 50 percent improvement over linear reference pipelines. He credited much of that to how cleanly the reads resolve homopolymers and tandem repeats.
Other Genomics News
Swedish private equity firm EQT is in “advanced talks” to buy Qiagen, Bloomberg News reported Oct. 20 (paywall). Shares of Qiagen are up 5 percent over the past five days trading on the New York Stock Exchange. Qiagen, which is based in Europe and offers assay kits and software to support both sequencing- and PCR-based molecular testing, has a market capitalization of €8.57 billion ($9.93 billion,) as reported by Reuters.
Myriad Genetics to Sell Prenatal Testing Business to Labcorp for $210M
Myriad Genetics has agreed to sell its prenatal health testing business to Labcorp for $210 million in cash.
Nvidia and the University of Washington’s Institute for Protein Design on Oct. 21 introduced RFD4-Proteina, a new open model for designing proteins and other biomolecules.
The model will be part of Nvidia’s BioNeMo platform, combining programmable protein design with search capabilities. It generates targeted designs that scientists and AI agents can refine. Potential applications include proteins designed to interrupt disease processes or detect signs of disease. Nvidia said the model is also available through partners including Amazon Bio Discovery, Benchling, Databricks, Google Cloud, Hugging Face, Nebius, OpenAI and Tamarind Bio.
Abbott's Cologuard Plus Receives European CE Mark
Abbott’s Cologuard Plus test has received CE Mark under the EU’s In Vitro Diagnostic Medical Devices Regulation (IVDR).
Takara Bio, SPT Labtech Partner on Automated NGS Library Prep
Takara Bio and SPT Labtech have partnered on validating automated library preparation workflows that pair Takara Bio’s next-generation sequencing library preparation assays with SPT Labtech’s Firefly liquid handling platform.
Element Biosciences kicked off a global road show Oct. 20 in Seoul, as reported by The Korea Herald. The company is now shipping its Cytos single-cell multiomics assay. Other stops on the tour will include Munich, Boston, Stockholm, and the company’s hometown of San Diego.
GeneDx announced Oct. 20 that it plans to start an early-access program that will provide long-read whole genome sequencing for certain patients. In a statement, the diagnostics company said it “will identify where long-read sequencing can deliver the greatest diagnostic value and offer [it] upfront for eligible patients.” The early-access program will start by the end of 2026.
Thermo Fisher Scientific saw third quarter revenue rise 9 percent, year over year to $12.14 billion. Academic and government revenue was flat.
Watchmaker Genomics has launched Expanse Whole-Genome Sequencing, a new PCR-free WGS assay that integrates library preparation and normalization into a single workflow. It uses the company’s EquiPlex Cas9-based normalization technology to enrich sequenceable molecules, producing a sequencing-ready library in approximately 2.5 hours from inputs as low as 75 ng, including from FFPE samples.
The company is launching an early access program now and will be sharing early performance data at ASHG. The company also announced Oct. 5 that the US District Court for the District of Colorado has granted its motion to dismiss patent infringement claims brought by Biomodal, which makes a competing methylation sequencing analysis method.
Tempus AI subsidiary Ambry Genetics has launched GenomeNext and GenomeReveal, whole-genome sequencing tests for patients with rare diseases. GenomeNext uses a pangenome reference and GenomeReveal adds RNA analysis to help interpret the findings.
“Advances in the use of a pangenome reference representing wider human diversity, like RNA analysis, and ongoing reanalysis are not just technical upgrades,” Wendy Chung, Chief of the Department of Pediatrics at Boston Children’s Hospital, said in a statement. “They lead to a more complete approach to clinical genomics that enables better variant detection, improves diagnostic yield across diverse populations, and provides clinically meaningful insights that evolve over time.”
Benchling has launched Science Foundries, a feature that lets R&D teams run an entire scientific process, such as lead optimization, assay development, or nominating a development candidate, with teams of AI agents working alongside scientists. Teams choose a process or goal and set which steps agents run and when a scientist must intervene. The agents then execute and monitor the work, coordinate handoffs and raise key decisions.
“Science Foundries in Benchling are the perfect complement to automated labs,” John McGee, cofounder of Parabilis Medicines, a clinical-stage biopharma, that has developed Foundries for protein production, screening and lead optimization, said in a statement.
Benchling said it is also building Foundries for therapeutic area research, preclinical studies and process development.
Freenome has launched the SimpleScreen Lung blood-based test, which allows eligible patients to be screened for both lung and colorectal cancer at the same time.
My Omics Reading List
Atanur, et al. “A phenome-wide structural variant map from 467,152 UK Biobank genomes.” Cell. Oct. 15, 2026.
An international team led by researchers from AstraZeneca looked at 1.6 million structural variants (SVs) “identified in whole-genome sequencing data from 467,152 UK Biobank participants, across 13,322 binary, 2,018 quantitative, and 2,923 plasma protein traits,” detecting 28,184 associations with human health outcomes.
“This study underscores the underestimated role of SVs in human genetics and their potential to advance understanding of human biology and drug discovery,” the authors wrote.
Elsewhere on Substack
Are biologists just hungry? James Hadfield goes over the litany of food-based metaphors for how genomics technologies work at his Core Genomics Substack.
Speaking of food, given where ASHG is located this year, here are some recs I’ve seen for local specialities:
Schwartz Delicatessen (Montreal-style smoked meat, similar to pastrami)
Van Horne Bagels (Montreal-style bagels)
Au Kouign-Amann (French patisserie)