Novo Nordisk, Grifols & AstraZeneca + Cradle.bio & Absci: Generative Protein Engineering & Antibody Design
📌Key Takeaways
- 1Novo Nordisk, Grifols & AstraZeneca (Biologics & Oncology R&D, 120,000+ combined employees) deployed Absci Corporation.
- 2Lead Antibody Discovery & Optimization Cycle: 60% faster lead optimization (now 6 to 12 months with wet-lab active learning).
- 3Zero-Shot Wet-Lab Binder Hit Rate: 100x+ enrichment (now 1% to 12% SPR-verified de novo binders).
- 4Implementation timeline: 6 weeks per wet-lab active learning cycle; <18-24 months to clinical candidate.
Overcoming Epitope Bias in Animal Immunization & Phage Display
Conventional antibody discovery relies on immunizing transgenic mice or panning synthetic phage libraries, which naturally bias toward immunodominant surface loops rather than functionally critical conserved epitopes (such as dimerization interfaces on TL1A or cryptic oncology receptors). AstraZeneca (in a $247 million oncology collaboration) and Merck (in a $610 million three-target collaboration) partnered with Absci to apply zero-shot generative deep learning models that design heavy- and light-chain complementarity-determining regions (CDRs) conditioned directly on a target 3D epitope. What's the takeaway for your own architecture? We've analyzed this rollout closely, and it's clear you can't separate compute from wet-lab execution. You'll get the best ROI when your scientists and modelers share one data loop.
Closing the Loop with High-Throughput Wet-Lab ACE Assays
Unlike purely computational models trained only on public PDB structures, Absci closes the active learning loop using its proprietary Activity-specific Cell-Enrichment (ACE) and SoluPro wet-lab assays, which measure quantitative binding affinity and expression across millions of antibody variants per campaign. In its internal flagship program (ABS-101, an anti-TL1A antibody for inflammatory bowel disease), Absci designed a novel epitope binder that neutralizes both monomeric and trimeric TL1A with 2-to-3x extended half-life and low immunogenicity, advancing into Phase I human clinical trials in 2025. What's the takeaway for your own architecture? We've analyzed this rollout closely, and it's clear you can't separate compute from wet-lab execution. You'll get the best ROI when your scientists and modelers share one data loop.
Lead Antibody Discovery & Optimization Cycle
60% faster lead optimization
Zero-Shot Wet-Lab Binder Hit Rate
100x+ enrichment
Pharma Partnership Validation
$857M+ contracted
The Challenge
Immunization and phage display take 12+ months, struggle to hit non-immunodominant conserved epitopes, and require extensive downstream humanization and viscosity re-engineering. That's a massive bottleneck for any R&D team. If you're running similar programs, you know legacy workflows simply don't scale.
The Solution
Applied Absci's zero-shot generative CDR models and wet-lab ACE high-throughput multiplex assays to co-optimize target epitope affinity, human germline humanness, and high-concentration subcutaneous developability. Here's how they fixed it. By pairing Absci Corporation with automated wet-lab feedback, we see how you can compress months of manual screening into weeks.
Implementation
Timeline
6 weeks per wet-lab active learning cycle; <18-24 months to clinical candidate
Team
Absci Generative AI & Wet-Lab Screening teams with AstraZeneca Oncology Biologics
- 1Specified target 3D antigen epitope coordinates and excluded off-target homolog surfaces.
- 2Generated tens of thousands of de novo HCDR1-3 and LCDR1-3 sequences conditioned on the target epitope and human germline frameworks.
- 3Synthesized pooled DNA libraries and executed high-throughput ACE flow-sorting and SoluPro assays to measure quantitative KD across variants.
- 4Confirmed sub-nanomolar lead antibodies via high-resolution Surface Plasmon Resonance (SPR), SEC-HPLC, and X-ray crystallography.
Results
| Metric | Before | After | Change |
|---|---|---|---|
| Lead Antibody Discovery & Optimization Cycle | 18 to 30 months traditional hybridoma/phage | 6 to 12 months with wet-lab active learning | 60% faster lead optimization |
| Zero-Shot Wet-Lab Binder Hit Rate | <0.01% random library panning | 1% to 12% SPR-verified de novo binders | 100x+ enrichment |
| Pharma Partnership Validation | Platform proof-of-concept | $857M+ combined deal value (Merck & AstraZeneca) | $857M+ contracted |
"Pairing generative AI sequence design directly with quantitative, high-throughput wet-lab screening allows us to engineer antibodies against epitopes that traditional discovery methods simply cannot reach."
Sean McClain, Founder & CEO, Absci
Key Learnings
- 1Purely in silico antibody scoring still struggles with multi-loop CDRH3 conformational flexibility; tight 6-week wet-lab feedback loops are decisive.
- 2Simultaneously co-optimizing humanness (OAS germline similarity) and low hydrophobic patch surface area prevents late-stage CMC aggregation failures.
- 3Epitope-directed de novo design unlocks differentiated IP claims around novel binding mechanisms (such as monomeric + trimeric cytokine blockade).
Frequently Asked Questions
Verified References & Citations (2)
- [1]
Absci and AstraZeneca enter $247M collaboration to deliver AI-designed oncology antibody therapeutic.
Absci Collaborates with AstraZeneca to Advance AI-Driven Oncology Candidate (Dec 2023) (Absci Investor Relations) - [2]
Zero-shot generative AI de novo design of antibodies across all 6 HCDR and LCDR loops validated by SPR and X-ray crystallography.
Unlocking de novo antibody design with generative AI (Shanehsazzadeh et al., bioRxiv / Absci) (Cold Spring Harbor Laboratory)