Enhancing ML-based binder design with high-throughput screening: a comparison of mRNA and yeast display technologies

Area of Science:

• Biotechnology
• Computational Biology
• Protein Engineering

Background:

• Machine learning (ML) accelerates in silico protein binder design, but experimental validation lags, creating a bottleneck.
• Miniproteins are promising therapeutics due to their size, stability, and production efficiency.
• High-throughput screening platforms are crucial for validating computationally designed binders.

Purpose of the Study:

• To compare mRNA display and yeast surface display for screening ML-designed miniprotein binders.
• To evaluate the performance of both platforms in identifying functional binders against TLT-1 and B7-H3.
• To assess the suitability of these platforms for therapeutic miniprotein discovery.

Main Methods:

• Screening of ML-designed miniprotein libraries using both mRNA display and yeast surface display.
• Direct comparison of platform performance using a shared DNA library.
• Biophysical characterization of selected binders to assess binding affinity and stability.

Main Results:

• Both mRNA and yeast display identified functional miniprotein binders.
• mRNA display demonstrated superior library coverage and preferentially enriched binders with slower dissociation rates.
• Selected binders from both platforms exhibited strong binding affinities and high thermal stabilities.

Conclusions:

• Integrating ML-based design with in vitro selection technologies like mRNA display offers a scalable framework for therapeutic miniprotein discovery.
• mRNA display is a powerful tool for high-throughput screening, potentially rescuing functional designs missed by cell-based systems.
• This study provides a validated approach for accelerating the discovery of novel miniprotein therapeutics.