Related Experiment Video
Updated: Nov 3, 2025
08:07
Probing High-density Functional Protein Microarrays to Detect Protein-protein Interactions
Published on: August 2, 2015
8.2K
High-throughput developability assays enable library-scale identification of producible protein scaffold variants
Alexander W Golinski1, Katelynn M Mischler1, Sidharth Laxminarayan1
1Department of Chemical Engineering and Materials Science, University of Minnesota, Minneapolis, MN 55455.
Summary
High-throughput assays predict protein developability, accelerating the discovery of therapeutic proteins. Machine learning models enhance prediction accuracy, streamlining protein engineering for biotechnological applications.
Area of Science:
- Biotechnology
- Protein Engineering
- Computational Biology
Background:
- Protein developability (expression, solubility, stability) is crucial for biotechnological applications.
- Traditional developability assays are low-throughput, creating a bottleneck in protein development.
Purpose of the Study:
- To evaluate high-throughput assays for predicting bacterial recombinant protein expression.
- To identify optimal assay combinations and leverage machine learning for enhanced predictive utility.
Main Methods:
- Deep sequencing of 10^5 protein variants sorted by proxied developability.
- Cross-validation and feature selection to identify informative assay combinations.
- Random forest model trained on assay performance to predict expression.
Main Results:
- Identified the most informative combination of high-throughput developability assays.
- Machine learning models significantly improved prediction accuracy (35% closer to experimental variance).
- Developed a model that trains 80% more efficiently than sequence-based models.
Conclusions:
- Validated high-throughput assays enable large-scale identification of developable proteins.
- Machine learning enhances the predictive power of developability assays.
- Accelerated protein commercialization by reducing development bottlenecks.
