That’s what Stanford computer scientist and computational biologist Jure Leskovec said he wanted when he started work on Biomni, a “general-purpose biomedical AI agent” formally introduced yesterday in Science.

“Most AI models in biology today are very specialized: for one specific task, for one specific data type, for one specific prediction,” he said in a press conference associated with the paper’s publication. “They are very powerful, but they are kind of locked into a single problem, and Biomni is not. [It] sits on top of these specialized tools and can execute long-range multi-step tasks across the whole landscape of biomedicine, from genetics to genomics, immunology, pharmacology, microbiology, and so on.”

Started in 2024 and put up as an open-source project in 2025, Biomni appears to be the first such agentic AI research tool. It’s already being commercialized by a startup, Phylo, and heralds a new wave of technologies standing on the shoulders of the large language models behind products like Google Gemini and Anthropic’s Claude.

Just last week, Anthropic launched Claude Science, another wide-ranging research aide that is also based on AI agents. And a major leap in Claude’s ability to write code means even more are likely on the way.

Researchers are already lining up to try them out. According to Leskovec and his Phylo cofounders Kexin Huang and Jerry Qu, more than 10,000 people have already started using Biomni. Whether they’ll choose to use it to augment the work of lower-level researchers or use it to replace them remains to be seen.

Leskovec is hopeful it will let trainees move off tedious work and “[let] the junior scientist get to the interesting, judgment-heavy part of the work faster,” he said. “Because Biomni is general, it means that any junior scientist now has access to this seasoned collaborator who is also an expert in, and knows how to use, tools from other disciplines or sub-disciplines.”

My guest this week, Johnny Yu, has seen firsthand how the new code-writing capabilities of LLMs has opened new horizons for the use of AI in biology. Tahoe Bio, a company he cofounded, has built its own AI agents, which are helping it rethink how to use AI in drug discovery and development.

Join us to learn more about what AI agents are, how they’re made, how they’re changing discussions about the impact AI can make in biology, and the limitations they’ll still face.

Subscribe now

You can also listen to the interview on Apple Podcasts and Spotify:

Show notes

Tahoe Bio graphic on the difference between biology models and reasoning models

Claude science blog post https://www.anthropic.com/news/claude-science-ai-workbench

Phylo: https://phylo.bio/about

LinkedIn post from Johnny on Rhaister and Tara:

Blog post from Tahoe Bio CEO and Cofounder Nima Alidoust:

Sullivan comes from Caris Life Sciences, where he served as chief commercial officer overseeing global commercial operations across oncology diagnostics and precision medicine. Before Caris, he held senior commercial leadership roles at Roche Diagnostics, Ortho Clinical Diagnostics, IDEXX Laboratories and Abbott Diagnostics.

“[Sullivan] brings deep commercial expertise, a customer-first mindset, and significant experience scaling global organizations,” Illumina CEO Jacob Thaysen said in a statement.

He also brings decades of experience in diagnostics, a major driver in the growth of clinical sequencing, now Illumina’s largest end market.

Subscribe now

Coletti’s appointment is effective Aug. 3. She will lead the company’s global legal, regulatory, and government affairs teams as well as serve as corporate secretary to Illumina’s board. She joins from Align Technology, where she was chief legal and regulatory officer. Before that, she held senior legal leadership roles at Danaher and Bayer HealthCare. She also serves as a director for Fortis Life Sciences.

The IVDR replaces the EU’s previous In Vitro Diagnostic Medical Devices Directive and requires more extensive evidence of analytical and clinical validity, plus quality system management, than the outgoing rules. Natera said the certification covered the full Signatera platform — the assay, specimen collection kit and associated software — and will let the company continue offering the test to EU patients after the prior directive’s transition deadline in 2028.

Under the certification, Signatera is indicated for adjuvant and surveillance monitoring across gastrointestinal, genitourinary, lung, head and neck, breast, skin and gynecological cancers, as well as diffuse large B-cell lymphoma, indolent non-Hodgkin’s lymphomas and pan-cancer immunotherapy monitoring.

In a statement, Natera said its test is the first personalized MRD test for solid tumors to clear that framework in the EU.

The certification follows two other recent regulatory decisions for Signatera: Japan’s Pharmaceuticals and Medical Devices Agency approved the test for colorectal cancer patients in June, and the FDA approved Signatera CDx as a companion diagnostic for muscle-invasive bladder cancer in May.

“MRD testing is redefining how we assess recurrence risk and guide treatment decisions for patients with cancer,” Julien Taieb, head of gastroenterology and gastrointestinal oncology at Université Paris-Cité, said in a statement. “This certification for Signatera is an important milestone as it will enhance access to personalized MRD testing for patients across Europe within a more rigorous regulatory framework.”

The updated test showed 80 percent sensitivity for detecting colorectal cancer, including 52 percent of Stage I cases, and 100 percent of Stage II, 97 percent of Stage III and 100 percent of Stage IV cases. Sensitivity for advanced precancerous lesions (APLs) rose to 18 percent from 14 percent in the first-generation test, and sensitivity for APLs with high-grade dysplasia rose to 42 percent from 31 percent. Specificity for no findings on colonoscopy was 90 percent.

The results trigger a $70 million milestone payment under a commercialization deal with Abbott. However, that’s short of the $100 million maximum payment, according to a July 9 research note from Canaccord Genuity Analyst Kyle Mikson, “since SimpleScreen v2 did not completely meet the specified thresholds of ≥=19 percent APL sensitivity and ≥83 percent overall CRC sensitivity.”

The study readout is positive for SimpleScreen but does not suggest it will not yet be a major threat to Guardant Health’s competing Shield test. Mikson said SimpleScreen v2’s roughly 80 percent CRC sensitivity and 90 percent specificity compared with about 84 percent sensitivity for Shield v2 at a similar specificity, though SimpleScreen’s precancer detection was notably stronger. “Shield will remain the leading blood test for CRC screening going forward,” he wrote.

Freenome submitted a premarket approval application to the FDA for the first-generation SimpleScreen CRC test in August 2025; the agency’s review is expected to complete in mid-2026. The company plans to submit a supplemental PMA for the updated version.

The validation used blinded, previously unevaluated samples from the PREEMPT CRC study, a prospective registrational trial that enrolled 48,995 asymptomatic, average-risk adults ages 45 to 85 at more than 200 sites. The analysis included 89 individuals with colorectal cancer, 1,570 with advanced precancerous lesions, and 157 with APLs showing high-grade dysplasia.

Under a commercial collaboration Freenome signed with Abbott in August 2025, Abbott will exclusively commercialize SimpleScreen CRC, once it’s approved. That agreement ties a milestone payment to this study’s results; based on today’s data, the milestone will be set at $70 million rather than the $100 million maximum, since the updated test did not fully meet the study’s highest performance thresholds.

“These results strengthen our confidence in the role blood-based colorectal cancer screening can play in expanding screening,” Jake Orville, senior VP of Abbott’s cancer diagnostics business, said in a statement. “Combined with Abbott’s preferred, guideline-supported screening option, we’re uniquely positioned to offer healthcare providers and patients a differentiated colorectal cancer screening portfolio built on innovation, performance and choice.”

Freenome, which is privately held, is planning to go public through a merger with Perceptive Capital Solutions, a special-purpose acquisition company (SPAC); a shareholder vote on that deal has been postponed to July 15 from July 9, according to Mikson.

Under a supply agreement announced July 8, Bifrost Biosystems will integrate Illumina’s sequencing-by-synthesis chemistry into its optical pooled screening (OPS) platform. Illumina’s reagents will be packaged and distributed as part of Bifrost’s OPS-optimized reagent kits, supporting in situ genotyping workflows on Bifrost’s system.

The deal locks in commercial access to Illumina’s sequencing chemistry, clearing one of the last technical hurdles standing between its cell-screening platform and a planned 2027 launch.

OPS is a technology that links complex cellular phenotypes, determined by imaging, to the genetic perturbations responsible for them, allowing researchers to study large numbers of genetic changes at once. It is immediately adjacent to spatial perturbation screening approaches, as discussed in the Ion Genomics podcast from June 5, 2026.

Bifrost, a Berkeley, California-based company with approximately 10 employees, licensed its core technology from the Broad Institute in 2024. On its website, Bifrost says it is a spinout from Johan Paulsson’s Harvard lab, but notes that Broad’s Paul Blainey and Harvard University’s George Church are also cofounders.

Blainey has also done key work in optical pooled screening, publishing a preprint on the method OttoSeq in April. Sami Farhi, a Broad researcher who also runs the spatial core lab there, told Ion Genomics that while his lab developed their own method for spatial perturb-seq “we use Paul [Blainey]’s method right now.” He’s using the approach to investigate how neuron morphology is affected by different psychiatric genes.

Genentech’s Aviv Regev and Eric Lubeck have also developed an OPS method, published a year ago in Nature Biotechnology.

The Illumina agreement is the company’s latest step toward commercializing the platform, which it said will support upcoming early-access collaborations with research institutions and other partners.

“This agreement represents an important validation of the maturity of the Bifrost platform,” Bifrost CEO Jonas Jarvius said in a statement.

Subscribe now

Ion Genomics Updates

AI lab agents — software that can autonomously plan and execute complex scientific tasks — are moving from concept to reality in the life sciences. From Anthropic’s Claude Science to homegrown tools at startups like Tahoe Bio, researchers are turning to these software intermediators to get work done. This Friday’s podcast breaks down what that means for genomics.

To help break it all down, I’ve turned to Johnny Yu, chief scientific officer of Tahoe Bio, a startup that has been building huge datasets to train AI models with the aim of improving drug discovery and development. His company recently came up with their own agent-based approach to modeling biology.

We’re also just over a week away from another happy hour on Friday, July 17, this time with the wonderful Molly Zeller, who runs a DNA sequencing core lab at the University of Wisconsin. Our first happy hour also had a Madison, Wisconsin connection, so maybe there’s something to draw on.

Make sure to sign up for a Substack account before the happy hour at 6 p.m. Eastern / 5 p.m. Central / 3 p.m. Pacific on the 17th to get access to the live broadcast, or catch it later on the usual podcast channels.

Other genomics news

23andMe Data Breach $47M Settlement Approved

A Missouri-based US bankruptcy judge has approved a $46.8 ​million settlement for victims of a 2023 data breach at ‌the genetic testing company 23andMe, as reported by Reuters on July 7. Approximately $14.3 million has already been disbursed.

The breach exposed genetic and other personal information of an estimated 6.9 million customers. California’s Attorney General has also filed a lawsuit related to the data breach.

Automated Genome Reanalysis Adds 5 Percent Yield

An automated process for continuous genome reanalysis developed by Australian researchers has been put into production, as reported by Yaakov Zinberg of GenomeWeb. I reported on the pipeline in 2024 before it had a name; now called Talos, the software made 241 diagnoses after being presented with 4,735 cases of unsolved rare disease, an additional yield of just over 5 percent.

In a paper published June 24 in Nature Medicine, researchers led by Zornitza Stark of the Murdoch Children’s Research Institute in Melbourne, Australia, suggested that nearly half were due to “improved analysis strategies. About a third stemmed from newly established gene-disease relationships and roughly one in five from new variant-level evidence.

Readers may recall that Illumina recently launched a study to apply both genome reanalysis and its new TruPath whole-genome assay to unresolved cases of childhood rare disease. It seeks an additional yield of more than 10 percent.

Simplify Genomics, a Human Longevity spinout, has signed a deal with SimonMed to integrate genome interpretation with whole-body MRI. Financial and other details were not disclosed.

What I’m Reading

Male runners may be twice as likely as female runners to “hit the wall” and slow down deep into a marathon, according to a recent study published in Scientific Reports.

The researchers analyzed 873,334 Berlin Marathon race times between 1999 and 2025. They found that while male runners tended to finish the marathon faster than female runners, the men were twice as likely to hit the wall and slow down by more than 20 percent. Only about a third of male runners did not slow down, while more than half of female runners were able to maintain an even pace.

The authors suggested that men may be going out too hard. Though I haven’t (yet) run a marathon, based on my college 8K races, this is a decent bet as to what’s happening. In fact, the faster a guy is, the more likely he is to hit the wall: among runners who finished in under three hours (or just under 7:00 per mile pace), “male runners were approximately six times more likely to experience catastrophic deceleration than their female counterparts,” the authors wrote.

My takeaway is that it’s a good idea to go out slightly slower than your goal pace for the first five or so minutes of a race. If you’re doing well you can make it up later and if you can’t, well, you probably got your pacing wrong to begin with.

Elsewhere on the Internet

The US Men’s National Team crapped the bed in their Round of 16 match against Belgium. I had previously said I thought the team could make the semifinals based on the way they were playing in early games. Unfortunately for me, they looked nothing like the swashbuckling team they started the tournament as.

Still, I’ll always have midfielder Malik Tillman’s magical free kick, the only US goal in the match. Belgium would retake the lead just over a minute later, so the joy was short lived, but what joy!

Under a supply agreement announced July 8, Bifrost will integrate Illumina’s sequencing-by-synthesis chemistry into its optical pooled screening (OPS) platform. Illumina’s reagents will be packaged and distributed as part of Bifrost’s OPS-optimized reagent kits, supporting in situ genotyping workflows on Bifrost’s system.

OPS is a technology that links complex cellular phenotypes to the genetic perturbations responsible for them, allowing researchers to study large numbers of genetic changes at once. It is closely related to spatial perturbation screening approaches, as discussed in the Ion Genomics podcast from June 5, 2026.

Bifrost, a Berkeley, California-based company with approximately 10 employees, is a spinout from Johan Paulsson’s Harvard lab and licensed its core OPS technology from the Broad Institute in 2024. Additional cofounders include Broad researcher Paul Blainey and Harvard research George Church. The Illumina agreement is the company’s latest step toward commercializing the platform, which it said will support upcoming early-access collaborations with research institutions and biotechnology partners.

“This agreement represents an important validation of the maturity of the Bifrost platform,” Bifrost CEO Jonas Jarvius said in a statement. “Securing commercial access to Illumina’s high-quality SBS chemistry enables us to deliver a fully integrated solution to our customers and accelerates our path toward widespread deployment of optical pooled screening.”

OncoAct covers the full diagnostic workflow, from raw whole-genome sequencing data to clinical reporting and therapy guidance. Hartwig says it is the first software to receive IVDR CE certification for whole-genome cancer diagnostics in solid tumors.

Announced July 2, “this certification confirms that […] OncoAct, meets the highest European standards for quality, safety and performance. It is an important recognition of the work we have done together with our clinical partners and, ultimately, a step forward for hospitals and patients who can benefit from more comprehensive cancer diagnostics,” Robert Jan Lamers, managing director of Hartwig Medical Foundation, said in a statement.

OncoAct’s analysis capabilities include genome-wide detection and reporting of somatic and germline variants, complex biomarkers including microsatellite instability and homologous recombination deficiency, high-resolution HLA typing, pharmacogenomics, virus detection, and tissue-of-origin prediction. Findings are automatically linked to standard-of-care treatments and relevant clinical trials.

Hartwig said the certification is intended to support broader hospital adoption of WGS-based molecular diagnostics.

As of the July 1 launch, the test is available as a laboratory-developed test for a self-pay rate of $3,500 per test, Guggenheim Securities Analyst Subbu Nambi wrote in a note to investors.

That’s much more than the two other MCED tests currently available as LDTs: Grail’s pathsetting Galleri and Exact Sciences’ Cancerguard (now part of Abbott) both currently run around $700 for self-pay customers.

Caris Detect draws on the company’s molecular profiling database and is supported by data from the Caris Detect ACHIEVE 1 study, which the company said demonstrated performance in detecting cancer signals across multiple cancer types. Caris did not disclose specific performance metrics from the study in its announcement. Nambi noted that Caris marketing materials showed Stage I sensitivity of 56.8 percent, stage II sensitivity of 70.1 percent, stage III sensitivity of 77.1 percent , and stage IV sensitivity of 99.1 percent; Stage I-II combined sensitivity is 61.3 percent.

How that compares to the other tests, namely Galleri, isn’t clear. The Caris test uses deep sequencing, likely boosting sensitivity, Nambi told Ion Genomics. “However, the data they have shared is not interventional or prospective which makes head to head comparison between MCED tests a bit difficult.”

For comparison, in May, Grail presented data from its NHS Galleri trial, suggesting that “annual testing with Galleri increased Stage I-II cancer diagnoses by 16 percent, when added to the standard of care.”

In its statement, Caris said it intends to continue building the clinical evidence base for the test.

“Early detection remains one of the most urgent and consequential challenges in oncology,” Caris CEO and Founder David Halbert said in a statement. “Caris Detect represents a major step forward in our mission to transform cancer care through the most advanced molecular science available. By combining unmatched biological depth with sophisticated AI, we are giving physicians and patients a powerful new tool to help identify cancer earlier and support more informed clinical decisions.”

Ion Genomics Update

Mark your calendars for Friday July 17! I’ll be doing another live happy hour, joined by the incomparable Molly Zeller, lab manager at University of Wisconsin-Madison DNA sequencing core facility. Two topics I can guarantee we’ll be talking about: sequencing technology and Wisconsin beer.

Make sure to sign up for a Substack account to get access to the live broadcast, or catch it later on the usual podcast channels.

Subscribe now

Other genomics news

Natera faces class action suit for discontinued fertility test

A class action suit filed in US federal court accuses Natera of false and misleading advertising for its discontinued preimplantation genetic test for aneuploidy (PGT-A.)

As reported by GenomeWeb, the suit alleges Natera “committed fraud by marketing and selling its Spectrum PGT-A test as a ‘proven, accurate, and reliable method to decrease the chance of miscarriage and increase the chance of giving birth to a healthy baby when science does not support this.’”

https://www.genomeweb.com/molecular-diagnostics/natera-faces-class-action-lawsuit-misleading-genetic-testing-claims

More Natera News

Samsung Affiliates Invest $110 million in Grail

Under an agreement announced June 25, Samsung-associated entities invested $110 million in Grail through the purchase of common stock at a price of $70.05 per share.

“This investment from the Samsung entities further strengthens our balance sheet and extends our cash runway as we advance key priorities, including securing regulatory approval and reimbursement for Galleri in the US and expanding access to MCED internationally,” Grail CFO Aaron Freidin, said in a statement.

As previously announced, Grail and Samsung intend to collaborate to commercialize the Galleri MCED test in South Korea, with the potential to expand into additional Asian markets, including Japan and Singapore.

Last month, Samsung poured $175 million into Element Biosciences.

Inso Biosciences Nabs $200,000 BARDA Funding

Cornell University spinout Inso Biosciences has received $200,000 in non-dilutive funding for technical development, evaluation, and validation of its sequencing sample prep technology. The funding is through Vanguard, a joint initiative between the BioTools Innovator startup accelerator and the US Department of Health and Human Services’ Biomedical Advanced Research and Development Authority (BARDA).

Inso has ties to the lab that developed zero-mode waveguides, the silicon chip features that power Pacific Biosciences’ long-read sequencing technology. The company is developing tech for purifying high-molecular weight DNA purification ahead of long-read sequencing.

NeoGenomics Appoints Carolyn Starrett to Board

Starrett, formerly CEO at Flatiron Health, brings more than 25 years of experience in health tech, data, and AI. She was previously on the board of Foundation Medicine and is currently a life sciences advisor to OpenAI.

CHOP Inks MOU With Abu Dhabi DoH

Children’s Hospital of Philadelphia (CHOP) announced the signing of a Memorandum of Understanding (MoU) with the Abu Dhabi Department of Health establishing a strategic collaboration focused on advancing gene editing capabilities and accelerating innovation in rare disease care.

Under the agreement, CHOP experts will provide specialised training to Abu Dhabi clinician-scientists in innovative gene editing approaches for rare diseases. The collaboration will focus on building local expertise and further refining specialised skills in base-editing therapies targeting conditions including liver metabolic disorders, genetic clotting disorders, urea cycle disorders and central nervous system conditions, while supporting the long-term development of advanced therapies for diseases prevalent within the region.

In 2025, CHOP created the first personalized gene‑editing therapy for a baby with a rare metabolic disease, https://www.chop.edu/news/childrens-hospital-philadelphia-marks-one-year-anniversary-worlds-first-personalized-crispr

Elsewhere on Substack: Springer Nature Sells Scientific American

What I’m Reading

As reported by GEN:

The first beat I had in genomics was CRISPR and genome editing technologies, so there’s a pang of loss in seeing that Sangamo Therapeutics has filed for bankruptcy. I recall that Sangamo’ Fydor Urnov was one of the most visible proponents of genome editing in those days, especially for the therapeutic potential of CRISPR.

Aimed Bio is also making an equity investment in Inocras for an undisclosed amount.

Under the agreement, announced July 1, AimedBio will apply Inocras’ whole-genome sequencing (WGS), multiomics, and data analysis capabilities across its ADC clinical trials to support biomarker identification, patient selection, and target discovery.

The two companies also plan to explore joint commercial opportunities, connecting AimedBio’s precision drug screening business with Inocras’ diagnostic platform.

In a statement, the companies said they aim to accelerate clinical development and uncover new opportunities for precision drug discovery.

“This partnership reflects what we’ve believed from the start, that cancer whole genome data has a role beyond diagnosing cancer. It can directly shape how new treatments are developed and identify which patients are most likely to benefit,” Inocras CEO Jehee Suh said in a statement. Finding likely responders is especially important “as ADC development grows more competitive,” AimedBio CEO Nam-Gu Her added.

AimedBio is a spin-off from South Korea’s Samsung Medical Center. Inocras is headquartered in San Diego, with other offices in South Korea and Hong Kong.

The companies declined to disclose a specific number of patient samples that Inocras will process under the agreement.

Under the terms of the agreement, announced June 30, Bioptimus will sponsor the generation of new spatial transcriptomics data on the FFCD’s patient cohorts. That data will measure and map gene activity within tissue samples, enabling researchers to use AI on routine pathology slides to predict gene expression data directly from the image.

In exchange, the FFCD gains access to Bioptimus’s M-Optimus multimodal foundation model. The FFCD retains full ownership of all generated data, which it will share with Bioptimus, and will have access to all newly produced spatial omics data for use in its own research programs.

Financial terms of the agreement were not disclosed.

The deal advances Bioptimus’s Spatial Tissue Embedding Learning Atlas (STELA), a multimodal tissue atlas linked to clinical data. The FFCD dataset is the largest GI cancer spatial omics atlas generated to date, the partners said in a statement.

“High-quality clinical data is the key to building a foundation AI model in biotechnology and this agreement will boost vital precision oncology research and patient health,” David Cahané, cofounder and managing director of Bioptimus, said in a statement. The FCCD data “anchors STELA’s GI cancer coverage in the kind of high-quality, outcome-linked clinical data that spatial atlases rarely include.”

Bioptimus seeks 100,000 samples to build STELA, with plans to integrate H&E imaging, spatial transcriptomics, and RNA-seq. Its M-Optimus model, announced in late 2025, was trained on a proprietary dataset including millions of patients and samples from more than 50 organs.

By using the model, “we aim to uncover novel features of tumor heterogeneity in GI cancers and gain deeper insights into the complex interactions between the tumor and its microenvironment” Pierre Laurent-Puig of the FFCD said in a statement. “Ultimately, these findings may help reveal hidden mechanisms driving treatment resistance and response, paving the way toward more precise and personalized therapeutic strategies for patients.”

Bioptimus has raised at least $76 million to date.

Under the agreement, Aveta will use serial Signatera testing to assess molecular residual disease (MRD) and treatment response across neoadjuvant, induction, and adjuvant settings throughout the trial and follow-up period. Signatera will function as a secondary endpoint in the study.

APG-157 is an oral immunotherapy that has received FDA Fast Track and Orphan Drug Designations for the indicated disease. The Phase 3 trial builds on previously reported Phase 2 data of APG-157 monotherapy, which showed evidence of tumor control, deep molecular responses, and event-free survival outcomes.

The trial expects to enroll more than 800 patients across North America, Europe, Asia-Pacific, and Australia, beginning in the second half of 2026. Financial and other terms of the partnership were not disclosed.

Just under a million people are diagnosed with head and neck cancer around the world each year.

“Patients with locally advanced head and neck cancer continue to face substantial risks of recurrence despite aggressive treatment,” Parag Mehta, CEO and founder of Bedford, Massachusetts-based Aveta Biomics, said in a statement. Incorporating Signatera testing into its clinical trial “will allow us to further validate the ctDNA findings observed in Phase 2 while generating molecular response data that will advance the understanding of treatment benefits for patients and strengthen the regulatory submission.”

The partnership extends Natera’s footprint in head and neck cancer research. The company recently announced results from the prospective Phase 2 SINERGY trial, which examined Signatera MRD-guided treatment in the same tumor type in 27 patients.

“The trial demonstrated a promising response rate and survival, with roughly three-quarters of patients experiencing treatment de-escalation guided by ctDNA dynamics,” Ari Rosenberg, a researcher at the University of Chicago and presenting author of the study, said in a February statement accompanying the presentation of results at the 2026 Multidisciplinary Head and Neck Cancers Symposium. “This led to a reduction in the number of chemotherapy cycles along with lower rates of high-grade toxicity compared with historical controls. With these results, there is great potential for ctDNA dynamics to optimize treatment.”

Sid, a tech entrepreneur who up until that point had been running his company, GitLab, went all in, doing as much testing and as many treatments as he could.

It seems to have worked. Moreover, Sid’s team thinks they’ve been able to figure out what went right, thanks to heaps of molecular data they’ve collected, including single-cell sequencing data.

“It’s very difficult, if not impossible, to do causative work in an N-of-1 study, but because we have this longitudinal sampling over the course of Sid’s case, we can start to build a mechanistic hypothesis,” said Jacob Stern, a single-cell data expert and veteran of 10x Genomics who found his way onto Sid’s team. “Which seems to be that there was a bunch of combination immunotherapy that helped to rev up the immune system over the course of 2024.”

Join Sid, Jacob, and me as we discuss the “maximal diagnostics” approach and how Sid plans to bring it to other patients who find themselves in a similar situation.

Subscribe now

You can also listen to the interview on Apple Podcasts and Spotify.

During our chat, Sid mentions that he got a treatment from a company started by a friend he made at the Y Combinator startup accelerator. That company is Shasqi, founded by José Oneto.

To see a timeline of Sid’s experience and all the data he and his team collected diagnosing and treating his disease check out osteosarc.com.

Sid has also written about his experience on his personal Substack:

Sid’s Substack

I’m going Founder Mode on my cancer

Elliot Hershberg’s article about my cancer journey gave language to something I’d been doing instinctively over the past year: managing my health in Founder Mode…

Read more

6 months ago · 57 likes · 16 comments · Sid Sijbrandij

For more information about Sid’s policy proposals, check out slide 17 of his embedded PowerPoint at https://sytse.com/cancer/

For more information about the cancer “playbooks” offered by the Sijbrandij Foundation’s Future of Cancer Care Today program, see: https://sijbrandijfoundation.org/fcct

More information about the companies Sid is investing in to bring aspects of his maximal diagnostics approach to market is at Evenone.ventures.

The program, announced June 25, is valued at up to $15 million and runs through 2027, with the potential to expand based on scientific progress.

Under the collaboration, the two companies will build a shared data foundation spanning cancer and autoimmune disease and apply Immunai’s single-cell AI platform to thousands of patient samples to identify patterns of T-cell dysfunction. Findings will be validated in Immunai’s wet lab and feed into drug discovery and development projects at Boehringer Ingelheim.

T-cell dysfunction is implicated in both cancer and autoimmune conditions, but research in the two areas has largely proceeded independently. The collaboration is designed to surface shared biological mechanisms and therapeutic targets that may not emerge from hypothesis-driven or siloed approaches.

“Cancer immunology and autoimmune diseases both involve T-cell dysfunction, but they have largely been explored separately,” Immunai CEO Noam Solomon said in a statement. “This collaboration brings together large-scale, clinically grounded data, translational science and functional validation to support broad target discovery across both fields. By taking an unbiased approach across thousands of patient samples, we hope to uncover biological insights and therapeutic opportunities that would otherwise remain hidden.”

The deal is the latest in a series of partnerships built around Immunai’s platform. In January, the company announced a collaboration with Bristol Myers Squibb to apply its technology across clinical development programs in oncology. In May, Immunai announced a third expansion of its multi-year collaboration with AstraZeneca.

“This transaction is an important milestone towards delivering on our mid- to long-term strategic agenda,” Kai Beckmann, Merck KgaA CEO, said in a statement. “Bio-Techne is an outstanding fit that directly supports our strategic direction focused on delivering cutting-edge products and solutions across the entire industry value chain – from lab customers to those manufacturing in the biotech and pharmaceutical industries.”

The deal would bring Merck a technology portfolio including proteins and immunoassay kits, as well as the Lunaphore spatial proteomics platform.

“In addition, Bio-Techne’s RNAscope and related in situ hybridization technologies would strengthen the capabilities of Merck in spatial biology and diagnostics,” the companies said in a statement.

Bio-Techne employs more than 3,000 people, approximately 2,300 of them in the US, and operates 34 global locations and 15 manufacturing facilities across North America, Europe, and China. The company reported revenue of more than US$1.22 billion in its fiscal year ended June 30, 2025.

Bio-Techne also holds a stake in Wilson Wolf Corporation, a manufacturer of cell culture devices. Bio-Techne expects to acquire the remaining ownership in Wilson Wolf immediately following the end of calendar year 2027.

The acquisition, which has been approved by Bio-Techne’s board of directors, is expected to close by late 2026 or early 2027, subject to regulatory approvals and Bio-Techne shareholder approval. The acquisition will be funded through a combination of existing cash and new debt, with Merck KGaA stating it intends to preserve its credit rating.

The offer represents a 36 percent premium to Bio-Techne’s share price.

Guggenheim Securities and JP Morgan are acting as financial advisors to Merck KGaA, with Sullivan & Cromwell serving as legal counsel. Goldman Sachs is acting as exclusive financial advisor to Bio-Techne, with Sidley Austin acting as legal counsel.

The Austin, Texas-based company expects to commercially launch Signatera for colorectal cancer in Japan by the end of the year, but has yet to determine final pricing.

Shares of Natera rose 10 percent in June 24 trading on the Nasdaq.

“This approval marks an important milestone for Japanese patients with colorectal cancer,” Takayuki Yoshino, president of the Japanese Society of Medical Oncology (JSMO,) and program director of the CIRCULATE-Japan project, said in a statement. “Clinicians and medical societies in Japan deeply value the strength of the clinical evidence on Signatera, demonstrating its ability to inform treatment decisions.”

Commercialization will be supported by an existing position statement from the Japan Society of Clinical Oncology and guidance from JSMO, which recommends MRD testing in colorectal cancer.

More than 150,000 people in Japan are diagnosed with colorectal cancer each year, making it one of the country’s most common cancers. Signatera can help identify which post-surgical patients are likely to benefit from adjuvant chemotherapy.

Natera Chief Medical Officer Alexey Aleshin said in a statement that the company plans to file for additional cancer indications in Japan following the colorectal cancer launch, “with muscle-invasive bladder cancer (MIBC) representing our next planned submission in Japan.”

Earlier this week, the US National Comprehensive Cancer Network gave Signatera its top-level recommendation for use in MIBC. In May, the FDA approved the test as a companion diagnostic for the disease.

Last week at the annual meeting of the European Society of Human Genetics, Illumina announced the Beyond GiICS (Genomics in the Infant Intensive Care Setting) study, which has partnered with “baby animal” sequencing projects across the region, to analyze approximately 400 unresolved cases.

Some cases will get reanalysis with Illumina-developed bioinformatics tools, including its DRAGEN pipelines and new AI variant interpretation tools; however, select cases will get re-sequenced using the TruPath assay, which provides information on larger genomic features, including repeats, structural variants, and larger insertions and deletions.

There are two ways of looking at this development. On the negative side: it’s a tacit admission that short-read genomes may be leaving out clinically important variants when diagnosing genetic disease.

In pediatric rare disease, speed and diagnostic yield are two of the most important factors and Illumina is getting squeezed on both fronts. Long-read approaches from Pacific Biosciences are getting to a point where they’re becoming first-line tests for suspected genetic diseases (outside the NICU) and researchers at Children’s Mercy Kansas City have already launched a pilot study similar to Beyond GiICS, where they’re evaluating cases that were not solved by rapid WGS.

Moreover, on the speed front, nanopore-based approaches from Oxford Nanopore Technologies and, more recently, Roche, may soon lead to same-day analysis, compared to two to three days for rapid short-read WGS.

After providing the technology that established the field of newborn sequencing, Illumina must prepare to fend off competition, something its doing across many sequencing applications.

On the positive side, if Illumina is correct in its hypothesis that TruPath can boost detection rates by an additional 10 percentage points, it could strengthen its position in the newborn sequencing market. That would translate to somewhere between 40 to 50 percent diagnostic yield, as rapid newborn sequencing studies have shown yields of around 30 to 40 percent.

This is not necessarily a big market, however, it has been a winner politically — many US states and other nations have launched “baby animal” NICU sequencing programs — and if WGS-based newborn screening ever become widespread, establishing a bulwark with TruPath could pay dividends.

Other countries participating in Beyond GiICS include Austria, Czechia, and Turkiye. Illumina is in the process of finalizing a deal to bring on another Western European country.

Illumina also added a seat to its board and appointed Daniel Skovronsky to fill it, effective immediately. He currently serves as Chief Scientific and Product Officer at Eli Lilly. Lilly acquired his company Avid Radiopharmaceuticals, in 2010. He has previously served on the board of Myriad Genetics. Skovronsky holds a MD-PhD from the University of Pennsylvania.

Subscribe now

Other genomics news

Guardant Health, TwinStrand Legal Battle Takes A Turn

Guardant Health was unsuccessful in a bid to overturn an $83 million jury award to TwinStrand Biosciences from a patent infringement case. Moreover, a federal judge added a 6 percent ongoing royalty.

However, in parallel legal action before the US Patent Office Patent Trial and Appeals Board (PTAB), Guardant was able to get all 23 claims of a TwinStrand patent invalidated. PTAB review of another TwinStrand patent is ongoing. Guardant will appeal the royalty, according to Guggenheim Securities Analyst Subbu Nambi. “Should appeal efforts fail, Guardant could implement workflow changes” that the company believes would allow it to avoid paying TwinStrand any royalties, she wrote.

Swedish Entrepreneur Gifts Millions to Support Prostate Research

The Karolinska Institute (KI) said it has received SEK 80 million ($8.3 million) from entrepreneur Björn Gysell to support prostate cancer research. In a statement, KI said it will use the funding to develop diagnostics and therapies. The donation will also finance the application of AI and advanced genetic analysis to better understand patients’ response to treatment.

In April, Gysell committed SEK 30 million over three years to support prostate cancer-related research at Uppsala University.

Elsewhere on Substack

Journalist William Herkewitz explores whether aid cuts have exacerbated the current Ebola epidemic in the Democratic Republic of Congo. There are technological challenges, such as a lack of effective vaccines for this particular strain, but many of the issues are human problems.

“It’s hard to pinpoint what isn’t going wrong,” he writes. (Full disclosure: Will is a dear friend of mine.)

The Rift Valley Dispatch

This Ebola Outbreak Is a Nightmare. Are Aid Cuts to Blame?

If you haven’t been following this day to day, let me give you the short version: the Ebola outbreak in eastern Democratic Republic of the Congo is totally out of control…

Read more

a month ago · 3 likes · The Rift Valley Dispatch

Erratum

In my podcast last week with Jen Dionne, a researcher at Stanford University and CEO and Cofounder of Pumpkinseed, I said that the US Men’s National Team failed to qualify for the 2022 World Cup. This was incorrect; the US had actually failed to qualify for the 2018 World Cup.

I’ll chalk this one up to the 2018 World Cup qualifying failure still looming large in my mind. I watched (on TV) the US lose 2-1 against Trinidad and Tobago to seal their fate and it seems like we’ve been clawing our way back to relevance since then.

And relevant we are again! The USMNT’s two wins to start this tournament has jolted me awake. I told Jen that on the eve of the World Cup, my enthusiasm was muted. I’d like to state for the record that I’m fully aboard the hype train. The team is playing a visually and emotionally attractive style. We have a modern coach using modern tactics and we finally have a striker who can finish in front of goal. I don’t care if this comes off as American naïveté (I’m usually more measured in my expectations) but I think they could make it all the way to the semifinals.

Elsewhere on the Internet (and also on Substack)

A Microsoft computer scientist has tackled questions of AI sentience in a delightfully absurd way: by showing how one could theoretically build a large language model (LLM) — the kind powering your favorite AI chatbot du jour — inside the classic computer strategy game Age of Empires II.

In a preprint titled “If AI is sentient, then so is Age of Empires II” (AOE2), Adrian de Wynter, also of the University of York, argued that the representation of an LLM affects its perceived properties.

It’s an interesting point, but let’s not lose sight of the real contribution here: making logic gates and “perceptrons” using goats, an in-game unit that provides food resources to players as they take their civilizations out from the dark ages and into more advanced states of conquest. (Not sure why he specifically went with goats, when it could have also been sheep, or even turkeys.)

Also, this is a sentence you don’t often see in academic papers:

Corollary 1 (AoE II is Turing-Complete). Let I be an instance of AoE II with two players p0,p1. Assume p0 has a market, a monk, a monastery, a relic, six villagers, and five farms; while p1 has a scout unit and only attacks p0’s buildings. Then if I has no time or size limits and the terrain allows for buildings everywhere, the game session in I is Turing-complete.

De Wynter goes into more detail about his paper and digital goatherding on his Substack. Unfortunately, I was unable to find any explanation as to why he chose AoE2 to make his point, rather than any other game with a custom map editor.

<TODO: insert title>: an AI newsletter by Adrian de Wynter

If LLMs Have Human-Like Attributes, Then So Does Age of Empires II

Would you believe ChatGPT if it tells you ‘I know how you feel?’, if instead of a chat pane it is a bunch of goats in Age of Empires II? What about MIT undergraduate students walking into 77 Mass Ave carrying a piece of paper symbolising a word…

Read more

a month ago · 12 likes · 4 comments · Adrian de Wynter

If it’s not clear by now, AoE2 is a personal favorite of mine. My brother and I used to play against each other all the time when we were teenagers and we rekindled our rivalry during the COVID-19 pandemic thanks to the remastered version. Like many players and real-life would-be conquerors, I never quite managed how to sustain a military campaign without letting my home base and economy fall to pieces. The in-game AI (not to be confused with De Wynter’s goat-gate features) on the hardest settings always seemed to get the better of me.

That’s all for this week, I have to go practice my build orders.

Specifically, the guidelines recommend Signatera-guided adjuvant atezolizumab, initiated at MRD-positivity within one year post-cystectomy in patients who have not previously received immune checkpoint inhibitor treatment. Category 1 is NCCN’s highest designation, reserved for recommendations backed by high-level evidence and uniform consensus.

“For the first time, NCCN has incorporated ctDNA-MRD testing into clinical decision-making following cystectomy,” Matthew Galsky, deputy director of the Mount Sinai Tisch Cancer Center, said in a statement. “These recommendations are supported by prospective phase 3 evidence showing that a ctDNA-guided approach, using Signatera, can help guide post-surgical treatment decisions.” It also marks the third NCCN guideline update to include circulating tumor DNA MRD testing, following previous additions in Merkel cell carcinoma and diffuse large B-cell lymphoma.

In May, the FDA approved Signatera CDx as a companion diagnostic to identify patients with MIBC who may benefit from adjuvant immunotherapy.

Under the agreement, announced June 19, Pirche will lead customer education efforts around using its TxPredictor platform alongside Natera’s Prospera test, a donor-derived cell-free DNA assay used to detect organ rejection after transplantation.

Financial terms of the collaboration were not disclosed.

“Integrating the predictive power of our TxPredictor platform with Natera’s Prospera test gives transplant teams an unprecedented toolkit to personalize immunosuppression therapies and ultimately protect graft longevity,” Thomas Klein, Pirche founder and CEO, said in a statement.

Berlin-based Pirche’s TxPredictor platform uses proprietary algorithms to analyze histocompatibility and generate immunological risk scores before transplantation. Natera’s Prospera test provides molecular surveillance after transplant by measuring dd-cfDNA levels in a patient’s blood as a marker of organ injury or rejection.

German neurodegenerative disease research center DZNE will spend approximately €6 million ($6.9 million) to build a database of genomic features influencing Alzheimer’s, Parkinson’s, Huntington’s and other neurodegenerative diseases.

The project, announced June 19, aims to sequence genomes of approximately 25,000 people over the next two years, using samples collected through ongoing DZNE studies, including clinical trials and the Rhineland Study, a large population genomics study.

Researchers hope to identify features that influence disease risk and progression, with the goal of opening new avenues in diagnostics, prevention, and therapy.

“In recent years, it has become increasingly clear that genetic factors play a key role in many neurodegenerative diseases. This data base is therefore intended to help identify features in the genome that influence disease risk and disease progression,” Joachim Schultze, scientific director of DZNE, said in a statement. These data could help “pave the way for new approaches in diagnostics, prevention, and therapy,” he added.

A technical focus of the project is to detect abnormal repeat sequences in the genome. DZNE said it will use long-read sequencing technologies to capture these repetitions more reliably than short-read methods. The database will also incorporate DNA methylation data.

Whether they will use sequencing from Pacific Biosciences or Oxford Nanopore Technologies, the two established long-read sequencing technology providers, isn’t yet clear. In an email, DZNE researcher Marc Beyer said that information is “currently still confidential and cannot be released.”

That’s the scale that Pumpkinseed CEO and Cofounder Jen Dionne thinks her company can reach for analyzing proteins. The technology, developed in her lab at Stanford University, uses Raman spectroscopy to fingerprint molecules and Edman degradation to achieve sequencing-by-subtraction of peptides: 30 amino acids at a time now, possibly up to 300, or full-length proteins, in the future.

The ability to capture 10 billion amino acids per day would represent several orders of magnitude better throughput than the best and fastest proteomics methods available today, whether via mass spectroscopy or other single-molecule technologies being developed by competitors like Quantum-Si or Nautilus.

Remarkably, that’s still a long way from reaching the data acquisition rate of state-of-the-art high-throughput sequencers, but it’s enough to start thinking about making the type of impact that NGS had on biological research.

Join us as we discuss the underlying physics behind Pumpkinseed’s technology, Jen’s journey from lab to C-suite, the applications they’re already pursuing, and her love of sports, including the World Cup.

Subscribe now

You can also find the episodes on Apple and Spotify podcast platforms:

For more information, check out pumpkinseed.bio or email Jen.

Links to papers discussed in the episode:

Hu et al. Rapid genetic screening with high quality factor metasurfaces. Nature Communications, July 26, 2023.

Zhang et al. From Genotype to Phenotype: Raman Spectroscopy and Machine Learning for Label-Free Single-Cell Analysis. ACS Nano, July 1, 2024.

Stiber et al. Dynamic, single-cell monitoring of CAR T cell identity and activation with Raman spectroscopy. BioRxiv, February 23, 2026.

Dolia et al. Metadynamics and Raman Spectroscopy for Glycan Structure–Spectrum Mapping. Journal of the American Chemical Society, June 17, 2026.