← Back to Home

UCB Partners with Cradle to Integrate AI Protein Design into Antibody Drug R&D

From identifying initial molecules to preparing for preclinical studies, UCB will continuously feed experimental results back into project-specific models to help optimize antibodies. The collaboration tests whether AI can fit into everyday R&D; no drug candidates or data validating efficiency gains have yet been disclosed.

By SURL BioNews

Having a computer propose antibody designs is only the starting point for drug discovery; researchers must then validate them individually and decide which molecules warrant further investment. Belgian biopharmaceutical company UCB announced a strategic collaboration with Cradle on November 4 to incorporate an AI protein design platform into selected antibody discovery projects, seeking to turn the back-and-forth between design and experimentation into an R&D process that continuously builds on experience.

According to the companies’ announcements, the collaboration covers hit generation, lead optimization, and the preparation needed to enter preclinical studies. It expands UCB’s existing digital antibody discovery and engineering program: Cradle’s platform will be combined with UCB’s own antibody discovery capabilities, computational tools, and disease biology research. The companies have not disclosed the specific targets, diseases, or candidate molecules included in the collaboration.

At the core of this workflow is the continuous feedback of laboratory results into the models. After researchers test molecules proposed by the platform, they will feed the results into models built for individual projects to inform the next round of designs. The companies call this “lab-in-the-loop”; the intent is for the models to adapt as project data accumulate, helping scientists weigh multiple properties of candidate molecules simultaneously.

Cradle said in its announcement that the platform integrates AI molecular engineering, learning from experimental data, and the joint optimization of multiple properties into a single system. CEO Stef van Grieken emphasized use across projects: enabling scientists to work on more types of molecules and more complex engineering problems through a shared platform. This extends the significance of the collaboration beyond individual designs to how the entire discovery team plans experiments, accumulates data, and selects candidate molecules.

However, the announcements describe the planned workflow and have not yet provided validation results from this collaboration. The companies have not disclosed the scale of the training data, the number of molecules tested, success rates compared with existing workflows, or how much R&D time might be saved. Finding higher-quality antibodies faster therefore remains a goal of the collaboration and cannot yet be regarded as a demonstrated effect.

RTTNews separately reported that the financial terms of the collaboration were not disclosed. At this stage, there are also no animal or human study results from the collaboration available for evaluation. Moving from model suggestions to a drug suitable for development still requires experimental evidence to be established step by step; even if candidate molecules complete preparation for preclinical studies, this does not support assumptions about efficacy or safety in humans, or the likelihood of obtaining regulatory approval.

References

  1. The Pharma Letter
  2. UCB
  3. Cradle
  4. RTTNews