👋 Hey, let’s get into it. Merck may be in line for a new Guinness World Record. Drum roll, please...the most expensive brunch ever. It just agreed to pay $400 million upfront for SPR2015, an oral cancer drug from a company called SciBrunch. Yes, really.

The drug is designed to block the KRAS G12D mutation in its active “on” state, but it has yet to enter human trials. Merck’s license could reach $2.13 billion if development and sales milestones are met. That’s a remarkable tab for a promising approach whose first clinical results are still to come.

🧠 FDA Approves AbbVie’s Juvmo for Parkinson’s Disease
The FDA approved Juvmo (tavapadon), a once-daily medicine for adults with Parkinson’s disease, either alone or alongside levodopa. It selectively stimulates D1/D5 dopamine receptors, giving physicians another way to manage symptoms. In a 26-week study of patients taking levodopa, useful “on” time without troublesome involuntary movements increased 1.7 hours with Juvmo versus 0.6 hours with placebo, a 1.1-hour difference. The drug came through AbbVie’s $8.7 billion Cerevel acquisition, so this approval puts something tangible on that deal’s receipt. It treats symptoms; it does not slow the underlying disease.

🤝 Lilly Taps InnoCare for Five Targets, Up to $3.35B
Lilly and China’s InnoCare struck a research and licensing deal covering up to five undisclosed drug targets. InnoCare could receive up to $100 million in combined upfront and near-term payments, roughly $3.25 billion in later milestones, and royalties. Those later dollars depend on progress; the companies did not break out a guaranteed upfront check. Unlike Merck’s single-asset KRAS license, this gives Lilly several shots at finding candidates. There are plenty of zeros in the potential deal value and fewer target names, so the science will come into focus as programs advance.

🏛️ FDA Nominee Faces Questions on Vaccines and Agency Independence
FDA commissioner nominee Heidi Overton had her Senate confirmation hearing last Thursday. Senators pressed her on vaccines, mifepristone, and political influence over agency decisions. Overton affirmed that the MMR vaccine is safe and effective and discussed speeding drug development; questions about how independently she would lead the FDA remained in focus. For drugmakers, regulatory predictability matters as much as a promise to move faster. Keep in mind, this was testimony, not a confirmation vote or a new FDA policy.

💉 Novo Targets Monthly and Quarterly Injections in $1.3B Nanexa Deal
Novo Nordisk wants space on the injection calendar. It licensed Nanexa’s PharmaShell technology for up to five peptide programs across obesity, type 2 diabetes, and other cardiometabolic diseases, aiming for monthly or even quarterly dosing. Nanexa could receive up to €1.165 billion ($1.3B) plus royalties, but the upfront amount was not disclosed. This is a delivery-platform bet: making molecules last longer could become a commercial advantage alongside efficacy and tolerability.

Reach 1,900+ scientists, founders, investors, operators, and other professionals across biotech and pharma.

We feature one exclusive sponsor per issue. If your company offers something relevant to our readers, reply to this email and let’s talk.

Alright, back to the newsletter.

💰 Genentech’s Earendil Bispecific Deal Could Top $1.5B
Genentech will pay Earendil $55 million upfront for oncology bispecific discovery; total potential payments exceed $1.5 billion if milestones are achieved.

🫁 Merck and Daiichi Withdraw Lung-Cancer ADC Filing
Merck and Daiichi Sankyo withdrew their filing in previously treated extensive-stage small-cell lung cancer after FDA discussions about insufficient evidence.

💉 FDA Approves Lilly’s Once-Weekly Basal Insulin
The FDA approved Onswik, Lilly’s once-weekly basal insulin for adults with type 2 diabetes, intensifying its rivalry with Novo’s weekly Awiqli.

✍️ Roche Signs Atavistik Deal Worth Up to $2B
Roche is paying Atavistik $70 million upfront to discover small-molecule drugs for cardiovascular, kidney, and metabolic diseases. The full amount is based on milestones hit.

👁️ Kodiak Shares Jump 178% on Phase 3 Eye-Drug Win
Kodiak’s wet AMD drugs matched Regeneron’s Eylea (aflibercept) on vision; 54% of Zenkuda patients reached 24-week dosing after loading and individualized retreatment.

🎯 Henlius-Amberstone T-Cell Engager Deal Offers Up to $440M Per Target
Henlius and Amberstone will develop tumor-activated T-cell engagers against up to two targets; the $440 million per-target figure consists of potential milestones.

⚠️ FDA Warning Letter Flags Empower’s GLP-1 Compounding and Sterility Practices
A FDA letter flagged Empower’s compounded GLP-1 products as apparent copies of approved drugs and identified shortcomings in sterile-production practices.

🧬 Kyverna Reports One-Year CAR-T Benefit in Stiff Person Syndrome
In a 26-patient, single-arm study, Kyverna reported sustained one-year responses to its CAR-T treatment for stiff person syndrome, including a 49% median timed-walk improvement.

Anthropic says Claude found a previously unrecognized enzyme system in bacteriophages, the viruses that infect bacteria. It calls the system array-associated reverse transcriptases, or ARTs. Reverse transcriptases copy RNA into DNA. The enzyme itself was already known; the new observation is a pattern of neighboring repeats and an accessory protein that may mean these pieces work together. It’s an intriguing biological lead, not a new gene-editing tool ready for a bench protocol.

Anthropic put AI agents to work searching more than 200,000 reverse transcriptases and producing hypotheses for scientists to inspect. The company says roughly 950 agents worked over 21 hours. Those numbers describe the scale of the search, not how well the proposed system works. Human researchers then tested candidates in the lab and showed expression of short RNAs associated with the system. This is a specific experimental result, but it leaves the central biological question unanswered: What is the system for?

The CRISPR comparison is tempting because repeated sequences near bacterial defense machinery once led to a transformative tool. Similar-looking genomic clues do not guarantee a similar function, let alone programmable editing. Researchers still need to establish what ARTs do naturally, whether the findings reproduce across examples, and whether anyone can direct the system to do something useful.

That’s where the practical AI story sits. Searching enormous sequence collections and ranking promising oddities could save scientists time, particularly when there are more plausible experiments than a lab can run. The real test is whether those rankings lead to reproducible biology that human teams might otherwise have missed.

Claude found a compelling candidate for the workbench. The enzyme has yet to supply its job description.

Want to hear what biotech Reddit thinks about the “discovery?” Scroll on down to the Community Vibes section.

This is the approximate gross amount ADARx raised in its upsized IPO, selling 26.25 million shares at $17 each. The RNA-interference developer plans to use part of the net proceeds to complete its Phase 3 onvuzosiran trial in hereditary angioedema and advance other candidates.

Here’s what biotech Redditors are talking about:

🧢 One Job Opening. An Entire Department’s Qualifications.
Redditors are ranting about job postings that want one person to clone, express and purify proteins, run Biacore, coordinate CROs, and bring CMC and regulatory experience. Sure, hiring managers love a versatile scientist. But at some point “wear many hats” becomes “please be the entire department.” If your ideal candidate needs three résumés stapled together, you may have written more than one job opening.

🤨 Claude Found an Enzyme. Scientists Have Questions.
We covered Claude’s enzyme finding in today’s Deep Dive. Over on biotech Reddit, the conversation turned to credit: How much belongs to the model, and how much to the scientists who chose what to test and then did the bench work? Some commenters also wondered how loudly to call it a discovery before anyone knows what the system does. The genomic finding is interesting. The argument over AI credit may be even more familiar to working scientists.

🧠 A Brain Atlas Maps 6.3 Million Single-Cell Profiles
A new brain research resource compares 6.3 million cell profiles from 1,494 deceased donors to explore molecular patterns across psychiatric and neurodegenerative diseases.

🏢 Lundbeck Plans a Boston Innovation Hub and U.S. Headquarters Move
Lundbeck plans a Greater Boston innovation hub and a 2027 move of its U.S. headquarters from Deerfield, Illinois.

📚 PubMed Adds a Tool to Trace Related Research
NIH’s Linked Discoveries tool maps connections among PubMed papers, helping scientists follow related genes, diseases, and chemicals without certifying any study’s conclusions.

🍄 BARDA Awards Scynexis Antifungal Contract Worth Up to $214M
Scynexis secured about $18.5 million initially from BARDA for investigational antifungal SCY-247; exercised options could bring the contract to roughly $214 million.

🤖 NVIDIA-Backed Basecamp Raises $140M for AI-Designed Therapies
Basecamp Research raised a $140 million Series C led by S32, with NVIDIA participating. Its EDEN models draw on biological data to help design therapeutic candidates, including in vivo cell therapies intended to engineer cells inside the body. The funding will support those models and advance Basecamp’s pipeline toward clinical development. The $140 million gives the startup room to build its AI platform and push its own medicines forward at the same time.

💊 September 29, 1982: The Tylenol Murders That Changed the Medicine Cabinet
On this day in 1982, the first victims of cyanide-tampered Tylenol capsules died in the Chicago area. Seven people died in all. The crisis prompted a recall of roughly 31 million bottles; FDA packaging requirements followed that November, with federal anti-tampering legislation in 1983. The seals we check on medicine bottles today are a visible part of that response and a reminder that product safety builds trust. No one has been charged with the murders, and the case remains unsolved. 😳

Thinking back to Claude’s enzyme discovery, I hope the scientists who did the bench work are ready for the hardest part: writing the contribution statement for roughly 950 AI agents.

“Agents 1–949: searched sequences. Agent 950: also searched sequences. Humans: went to the lab to see if any of this was real.”

That’s all for today. See you Thursday 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