Machine Learning for Biologics: Opportunities for Protein Engineering, Developability, and Formulation.

Area of Science:

• Biopharmaceutical development
• Protein engineering
• Computational biology

Background:

• Successful biologics require simultaneous optimization of activity and physicochemical properties (developability).
• The design space for protein sequences and buffer compositions is vast, posing optimization challenges.
• Current development timelines and costs for biologics are significant.

Purpose of the Study:

• To highlight emerging applications of machine learning (ML) in biologics discovery and development.
• To discuss the potential of ML in optimizing protein properties for enhanced activity and safety.
• To explore how ML can reduce development time and manufacturing costs.

Main Methods:

• Review of current ML applications in protein engineering, biophysical screening, and formulation.
• Analysis of ML's capability to extract insights from complex biological datasets.
• Discussion on reducing experimental efforts for achieving multiple quality targets using ML.

Main Results:

• ML shows significant potential to accelerate and improve the optimization of biologics.
• ML facilitates simultaneous optimization of multiple protein properties, including activity and developability.
• ML reduces the experimental burden in achieving desired quality targets for biologics.

Conclusions:

• Machine learning is a powerful tool for advancing biologics discovery and development.
• AI and ML can significantly enhance protein engineering, screening, and formulation processes.
• Future AI interventions are anticipated across various stages of biological development.