👋 Hey, let’s get into it. Talladega Nights must have been playing in the FDA break room recently, because the agency just issued the regulatory equivalent of Ricky Bobby’s “I wanna go fast.”

The agency opened applications for its Expedited IND Pilot, one piece of Operation TrialBlazer. It will select 8 to 10 sponsor-institution pairs to prepare INDs, submit pieces for rolling review, and coordinate trial-site activation.

The fast lane is only a pilot for now. But after watching promising drugs idle at the starting line, the FDA is finally putting its foot on the gas. ‘Cause if you ain’t first, you’re last...right?

🤖 TikTok Parent’s AI Drug Spinoff Reportedly Valued at $1.5B
TikTok’s parent company is moving into a very different kind of algorithm. Reuters reports that ByteDance’s spun-off AI drug unit, Anew Labs, raised $290 million at a valuation of about $1.5 billion, with ByteDance retaining 56%. Neither the company nor its investors confirmed the financing. The round would give Anew considerable fuel for biomolecular modeling, antibody design, and drug discovery. But capital and compute can accelerate experiments, not biology’s willingness to cooperate. The real milestone is whether those models produce candidates that survive laboratory and clinical validation.

💰 GSK Signs Up-to-$750M Deal for a Trispecific Myeloma Drug
GSK agreed to acquire global rights to Chimagen’s preclinical trispecific T-cell engager for multiple myeloma in a deal worth up to $750 million, including an undisclosed upfront payment and milestones. The therapy is designed to bind a T cell and two tumor-associated targets, essentially giving one antibody three assignments. That complexity could improve activity or tolerability, but the targets remain undisclosed and human testing is not expected until 2027. GSK is paying for a design hypothesis with portfolio potential.

🧬 Lilly Offers Up to $237M per Program for QurCan’s Genetic-Medicine Delivery
Lilly is teaming with QurCan to deliver nucleic-acid medicines to the central and peripheral nervous systems using polymer-lipid nanoparticles instead of viral vectors. QurCan will optimize the TERP delivery technology, while Lilly handles later research, development, and commercialization. The economics include an undisclosed upfront payment, a strategic investment, and up to $237 million in research support and milestones per program, plus royalties. The number of programs was not disclosed, so there is no honest total deal value.

🧠 Novartis Commits $125M to Sironax’s Brain-Delivery Platform
Novartis will pay Sironax $125 million following closing for full global rights to its blood-brain barrier delivery platform, exercising an option created last year. The technology is designed to carry large cargoes, including antibodies and gene therapies, into the brain, where nature employs a very selective bouncer. Sironax keeps rights to use the platform with selected assets, so this is not a takeover of the company. It is a wager on an enabling technology that could support multiple neuroscience programs, provided the delivery data eventually match the ambition.

Stop hoping the algorithm finds your audience. They’re already here.

Put your brand in front of 1,600+ scientists, founders, investors, operators, and other biotech and pharma professionals.

Interested in sponsoring an upcoming issue? Reply to this email and let’s talk.

Alright, back to the newsletter.

🤝 Sanofi Trades 20 Medicines for a Cheplapharm Stake
Sanofi plans to transfer 20 established medicines and three manufacturing sites to Cheplapharm in exchange for a 26.4% equity stake.

⚖️ BMS and Ono Sue Amgen Over an Opdivo Biosimilar
BMS and Ono sued Amgen over its proposed Opdivo biosimilar, alleging infringement of seven patents while the therapy remains under FDA review.

🧠 Novartis Ends ALS Program After Phase 2 Failure
Novartis ended its lifonebart development program in ALS after the 251-person ASTRALS Phase 2 trial missed its primary and secondary endpoints.

💊 Sling Raises $123M for Its Thyroid Eye Disease Pill
Sling raised a $123 million Series C to carry oral IGF-1R inhibitor linsitinib through its ongoing Phase 3 thyroid eye disease program.

🏭 Reckitt Plans Up to $600M in U.S. Investment
Reckitt plans up to $600 million for U.S. R&D and manufacturing, including its New Jersey hub and North Carolina OTC facility.

💉 Novo Exits Its Monthly GLP-1 Deal With Ascendis
Novo terminated its Ascendis collaboration, returning rights to a monthly semaglutide program that Ascendis plans to continue developing for obesity and diabetes.

📈 Electra Targets a $325M IPO
Electra proposed a roughly $325 million IPO at the range’s midpoint to fund ipsoprubart and other inflammatory-disease programs through clinical development.

☢️ Curium Wins FDA Approval for Novartis Competitor
The FDA approved Bexlutry, a radioligand therapy for adults with somatostatin receptor-positive neuroendocrine tumors, creating a new commercial alternative to Novartis’ Lutathera.

Pharma has a data-sharing problem. Every drugmaker wants better AI models. Almost none wants to hand rivals the private protein structures that could make those models better. Fair enough. “Trust us with the crown jewels” is not much of a collaboration strategy.

Apheris, five drugmakers, and Columbia’s AlQuraishi group tested another approach. AbbVie, Astex, BMS, J&J, and Takeda trained a protein-ligand co-folding model across 20,167 proprietary structures without pooling the underlying data in one place. Each company trained locally, then shared model updates rather than confidential structures. Think group project, except nobody emails the spreadsheet.

The resulting consortium model improved the share of high-quality predicted protein-ligand interfaces from 35.6% to 52.1% on 1,056 held-out structures, a gain of 16.5 percentage points. This is important because predicting how proteins and potential drugs fit together can guide discovery decisions before researchers spend more time and money in the lab.

The business case may be just as important as the benchmark. Drugmakers collectively hold richer structural data than any public repository, but competitive and intellectual-property concerns keep much of it locked away. Federated learning offers a possible bargain: contribute what the data teach without surrendering the data themselves.

Still, this is consortium-reported evidence, the resulting model remains private, and better predictions are not better medicines. Outside researchers also cannot fully reproduce a benchmark built on proprietary structures. Federated learning reduces one barrier to collaboration; it does not eliminate security risks, commercial mistrust, or biology.

Pharma may never share all its secrets, but it may be willing to share what those secrets teach.

This is the percentage of its workforce Sionna announced plans to cut after its earlier cystic-fibrosis trial disappointment. The company is narrowing development around SION-451 and SION-2222, extending its cash runway into the second half of 2029. Employees are paying the most immediate price.

Here’s what biotech Redditors are talking about:

🤔 What Do You Do When You’re Not Sciencing?
One Redditor who loves science admitted that turning it into a career made them want to disconnect after work, then asked the community what fills their off-hours instead. Answers included fiction novels, long-distance running, gardening, textile crafts, climbing, fashion, and metal concerts. One self-described cliché runs and reads sci-fi. Another likes gardening because the results are visible and occasionally edible. So, how do you clock out of science? Let us know. If enough people reply, we’ll publish the results in the next issue. I’ll start: beach volleyball.

😅 Before Taking That Startup Role, Check the Runway
A biotech worker is considering leaving a 10,000-person company to lead 12 people as an R&D director at a startup. Nice title. But the reported $6 million Series B? That gave Reddit pause. With salaries alone estimated at $2.5 million annually, commenters advised treading carefully. Before taking the leap, ask how much cash remains, how fast it’s burning, and what milestone comes next. A fancy title feels less fancy when the runway comes to an end.

🧠 Brain Implant Gives an Avatar Speech and Body Language
An experimental brain implant let two people with paralysis control an avatar’s speech and expressive body language in real time.

⚖️ Court Blocks Four-Year Limits for International Students and Scholars
A federal court temporarily blocked fixed four-year admission periods for certain international students and scholars, preserving duration-of-status rules while litigation continues.

🤖 Biotech Recruiter Sees Demand for Biology-and-AI Fluency
Recruiter Darren Nelson argues biotech increasingly needs workers fluent in both biology and AI, especially data quality, engineering, and model validation.

🧐 Brain Atlas Reveals 1,067 Overlooked Microproteins
Researchers identified 1,067 previously uncharacterized brain microproteins, creating a new resource for studying overlooked biology potentially relevant to Alzheimer’s disease.

🤖 RonanRx Pitches AI-Personalized Peptides and GLP-1s
YC-listed RonanRx is building an AI-assisted care platform around peptides and GLP-1s. The company says it connects patient information, clinician evaluation, prescribing, and medication delivery to support individualized treatment. RonanRx describes itself as a technology platform working with a separate pharmacy, with prescribing and dispensing decisions left to licensed professionals. The pitch is fewer handoffs across a fragmented experience.

🦷 September 17, 1683: Leeuwenhoek Meets the Residents Between Our Teeth
On this day in 1683, Antonie van Leeuwenhoek wrote to the Royal Society about microscopic “animalcules” he found in material scraped from teeth. He examined his own mouth and samples from others, including a man who reportedly never cleaned his teeth, because apparently seventeenth-century science did not believe in easy lab days. The moving organisms were bacteria, revealing an unseen world inside ordinary dental plaque. Leeuwenhoek turned someone else’s mouth gunk into scientific correspondence. And now, a message from your local dentist: Please floss. That’s all we ask.

Every tech giant seems determined to get into biotech, especially AI drug discovery. TikTok’s parent reportedly has a $1.5 billion AI drug spinoff, so perhaps biotech should return the favor.

Hear me out: a biotech dating app where you match by therapeutic area and willingness to relocate to Boston or San Francisco. The tagline: Where “good chemistry” comes with supporting data.

That’s all for today. See you Tuesday for the next issue. 👋

✍️ Today’s email was brought to you by Josh Martin. Connect with me on LinkedIn.

If you’re enjoying BioNucleus, share it with a friend or coworker who’d get value from it. We’re a small operation, and every share makes a bigger difference than you think.

Was this email forwarded to you? Sign up here 👇

Subscribe for free

Follow us on social and stay one step ahead