Related Experiment Video
Updated: Jul 14, 2026
04:24
An Open-Source Framework for Mass Calculation of Antibody-Based Therapeutic Molecules
Published on: June 16, 2023
Automation and Active Learning for the Multi-Objective Optimization of Antibody Formulations
D Christopher Radford1, Matthew Tamasi1, Elena Di Mare1
1Department of Biomedical Engineering, Rutgers, The State University of New Jersey, Piscataway, New Jersey, USA.
Advanced Science (Weinheim, Baden-Wurttemberg, Germany)
|July 13, 2026
Summary
Machine learning accelerates antibody bioformulation by predicting excipient effects. This high-throughput pipeline optimizes therapeutic protein formulations, improving stability and performance.
Area of Science:
- Biopharmaceutical Engineering
- Protein Chemistry
- Computational Biology
Background:
- Monoclonal antibodies are crucial therapeutics, typically manufactured as preformulated solutions.
- Bioformulation of complex proteins like antibodies presents significant engineering challenges due to intricate excipient interactions.
- Traditional formulation development is time- and material-intensive, struggling with multidimensional design spaces and non-linear relationships.
Purpose of the Study:
- To develop and validate a machine learning-driven pipeline for high-throughput antibody bioformulation.
- To model and predict the impact of Generally Recognized As Safe (GRAS) excipients on antibody formulation behavior.
- To integrate automated Design-Build-Test-Learn (DBTL) cycles for efficient formulation optimization.
Main Methods:
- Utilized a high-throughput discovery pipeline incorporating machine learning algorithms.
- Employed automation-assisted "on-demand" formulation to generate diverse antibody solutions.
- Integrated an iterative DBTL cycle for model-guided experimental design and refinement.
- Conducted biophysical characterization of formulated antibody solutions.
Main Results:
- Achieved accurate predictive models for antibody formulation behavior across multiple objectives.
- Successfully identified improved antibody formulations balancing competing design criteria.
- Demonstrated optimization of key formulation parameters including melting temperature, diffusivity, and high-concentration viscosity.
- Mapped the property-function landscape of excipient effects on antibody formulations.
Conclusions:
- The machine learning pipeline effectively guides antibody formulation development.
- This approach accelerates the optimization of biotherapeutics by managing complex design requirements.
- The validated technique enables efficient exploration of the formulation design space for improved therapeutic agents.
