Ning Leng

Open source advocator & community builder in Pharma clinical trial data science. Stats PhD.

[Ning Leng @ningleng]

Two students take the same exam. The teacher decides to estimate grades when 90% of the allotted time has passed. One student turned in their paper early and scored 70. The other is still working. The teacher marks the second student as a fail. Two FDA cases in 2026 ran into this problem. In April, AstraZeneca's camizestrant was voted down 6-3 at ODAC. In July, Osteal Therapeutics' NEXCHANGE KIT received a complete response letter. Both trials reported statistically significant data. Both were challenged on the same grounds: the treatment protocol created a timeline advantage that the endpoint design failed to separate from genuine treatment benefit. https://t.co/WdxS8fY3kC

[Ning Leng @ningleng]

The First Major CRO-AI Lab Deal: What ICON + Anthropic Tells Us About Where Clinical Trials Are Headed https://t.co/raVPQc01kX

[James Zou @james_y_zou]

Excited to introduce Paperclip Chat!🖇️

Now you can use #Paperclip directly in your browser for deep literature synthesis for free
⚡️ Each claim is grounded with inline citations
⚡️ Upload your own PDFs. They're automatically integrated into Paperclip's virtual filesystem

[Ning Leng @ningleng]

@MEPFuller 🤣

[Ning Leng @ningleng]

"More drugs per target likely means lower value per discovered asset."

@benjamineliu just wrote the most important AI pharma essay of the year on @a16z.

Key takeaway: AI will transform pharma unevenly. Discovery gets easier first. Toxicity prediction next. Clinical efficacy last — because the datasets simply don't exist yet.

Even with a perfect predictive model, regulators won't drop the requirement for prospective human trials. The bottleneck stays physical.

His framework: use predictive models to maximize PoS and rank assets, but don't bet the company on being better than the 30% industry baseline. Take enough well-underwritten shots. 10 Phase 2 programs at 30% = only 2.8% chance of zero winners.

The value isn't in discovering molecules anymore. It's in proving they work.

[Ning Leng @ningleng]

A group called FulcrumSec sat inside Novo's systems for 2+ months, stole 1.3TB — 24 internal AI models, a nanobody discovery engine, a patent-writing LLM, clinical trial patient data, pipeline info on 5 unreleased drugs. No custom malware. No zero-days. Just keys left in plain sight.
Pharma companies have quietly become software companies, but the security culture is still built around negotiating Veeva contracts and passing GxP audits.

[Ning Leng @ningleng]

Many of my Silicon Valley friends are stressed about layoffs right now. But there is SO much work to be done outside tech.

Novo Nordisk — the $300B company behind Ozempic — just got hacked because a developer hardcoded an Azure credential in client-facing JavaScript. Anyone could press F12 and find it.

https://t.co/0oIO6HPSxW

[Mark Lewis, MD, FASCO @marklewismd]

Cheers, chills, and a standing ovation when RASolute 302 showed unprecedented survival on daraxonrasib for patients with progressive pancreatic cancer

Seldom do you sense you’re witnessing a historic moment in cancer care but this feels like ras targeting has arrived

#ASCO26