Lilly's Secret Weapon for Finding Better Antibodies Isn't a Lab
Eli Lilly just plugged Twist Bioscience's massive antibody libraries into its TuneLab AI platform, building a data flywheel that gets smarter with every experiment. No splashy price tag, but the strategic implications are enormous.
The Drug Factory That Runs on Data
Imagine trying to find a needle in a haystack. Now imagine that haystack contains ten billion straws, and only a handful of them could cure a disease. That's roughly what antibody discovery looks like: sifting through enormous libraries of protein candidates, hoping to find the few that actually work.
Eli Lilly just decided to make that search a lot smarter. On September 16, the pharma giant announced a partnership with Twist Bioscience, a synthetic biology company that builds massive antibody libraries using silicon-based DNA synthesis. The collaboration plugs Twist's wet-lab data capabilities directly into Lilly's TuneLab, an AI-powered drug discovery platform that Lilly launched in September 2025.
The goal is deceptively simple: let AI models learn from real experimental data, so each round of testing makes the next round faster and sharper. Think of it like a recommendation algorithm, except instead of suggesting Netflix shows, it's predicting which antibody candidates are worth pursuing.
What TuneLab Actually Does
TuneLab isn't a single model. It's a platform that gives biotech partners access to Lilly's predictive AI models, built on years of proprietary research data covering drug safety, pharmacokinetics (how drugs move through the body), and early development decisions. Lilly has said those models were trained on data from hundreds of thousands of unique molecules.
The clever part is what happens next. TuneLab uses federated learning, a technique where partner companies can contribute their own data to improve the models without ever exposing their proprietary secrets. Everyone's data stays locked in their own vault, but the collective intelligence grows. It's like a potluck dinner where nobody shares their recipe, but the food keeps getting better.
TuneLab already connects with tools biotech teams actually use, including Benchling and Schrödinger's LiveDesign platform. Early partners included Circle Pharma and insitro, and the roster has been growing since launch.
Get tomorrow's biotech intelligence before your competitors.
Join thousands of biotech professionals who start their day with our free, daily briefing.
Why Twist Is the Perfect Sous Chef
Twist Bioscience brings something Lilly's AI models are hungry for: high-quality experimental data. Twist's antibody platform is built on precision DNA writing technology, and the company maintains a collection it calls the "Library of Libraries": over 15 synthetic antibody libraries, each containing up to ten billion antibody candidates.
Those libraries come in multiple formats (VHH, Fab, and scFv, which are different structural flavors of antibody fragments), and Twist designs them with manufacturability baked in from the start. The company removes molecular features known to cause production headaches downstream, which is a bigger deal than it sounds. Plenty of antibodies look great in a computer model but fall apart when you try to make them at scale.
Under the new agreement, TuneLab users can order antibody characterization services from Twist using preferred protocols and pricing. The data generated feeds back into TuneLab's models, including AbLab, Lilly's antibody developability prediction tool. Every experiment Twist runs potentially makes Lilly's AI a little bit smarter.
The Deal Nobody Priced
If you're looking for a splashy dollar figure, you won't find one here. Lilly and Twist disclosed no upfront payment, no milestone structure, no royalty terms, and no deal duration. This isn't a traditional pharma licensing deal with billion-dollar biobucks. It's a services integration: Twist becomes a preferred data provider inside Lilly's ecosystem.
That might sound underwhelming until you consider what it really means. Lilly is building a data flywheel for drug discovery. Every partner that uses TuneLab generates data. That data improves the models. Better models attract more partners. More partners generate more data. It's the same playbook that made companies like Google and Amazon dominant in their fields, except this flywheel is designed to spit out drug candidates instead of ad revenue.
Wall Street noticed. Twist's stock jumped on the announcement, with reports citing gains of roughly 3.6% to 5% depending on the time of day. But not everyone was popping champagne; skeptics pointed out that Twist still isn't profitable and already trades at a premium valuation.
Lilly's Bigger AI Chessboard
This Twist deal is just one piece of a much larger strategy. Over the past year, Lilly has been assembling an AI arsenal that would make a tech company jealous.
In October 2025, Lilly and NVIDIA announced plans to build what they called pharma's most powerful supercomputer. By January 2026, that expanded into a Co-Innovation AI Lab with up to $1 billion in talent, infrastructure, and computing power over five years. Separately, Lilly's deal with Chai Discovery targets AI-designed biologics, including novel antibodies against multiple targets. And the company's collaboration with Insilico Medicine, originally a modest research deal, ballooned into an agreement worth up to $2.75 billion focused on AI-enabled oral drugs.
The pattern is clear: Lilly isn't betting on one AI approach. It's building an entire ecosystem, controlling the compute layer while tapping external innovation at every stage of the discovery pipeline.
Why This Matters Beyond Lilly
The broader trend here is impossible to ignore. AI-driven antibody discovery is projected to grow at a 22.4% compound annual rate from 2025 to 2030, according to recent market research. The industry is moving away from isolated models toward full-stack workflows where computational predictions and lab experiments feed each other in continuous loops.
Developability (whether an antibody can actually be manufactured and delivered as a drug) is no longer an afterthought. It's becoming a design constraint from day one. Teams are screening for stability, aggregation risk, and manufacturability at the same time they're optimizing binding affinity. That's a fundamental shift in how biologics get built.
And the competitive moat isn't just about having the best algorithm anymore. It's about having the best data. Fragmented, low-quality experimental datasets remain the biggest bottleneck in the field. Companies that can solve the data problem, by building partnerships like the one Lilly just struck with Twist, will have a structural advantage over those still working with patchy training sets.
The Bottom Line
Lilly isn't just buying antibody services from Twist. It's feeding its AI flywheel. Every experiment, every data point, every partner interaction makes the platform stickier and the models smarter. The Twist deal might not have a headline-grabbing price tag, but it's the kind of quiet infrastructure play that compounds over time.
The real question isn't whether AI will reshape antibody discovery. That's already happening. The question is whether Lilly's platform-first approach will give it a lasting edge, or whether the next wave of AI-native biotechs will build something even better from scratch. Either way, the old model of throwing molecules at the wall and hoping something sticks is fading fast.
Deals & M&A5 min read
Roche Just Paid $36.5M for a Machine That Tests 300,000 Antibody Combos
Roche just signed a deal worth up to $1 billion with a three-year-old startup that can screen 300,000+ bispecific antibody combinations at once. The real story isn't the price tag; it's what it says about the future of drug discovery.
