Soluble protein analog selection engine (SPASE): An automated AI-powered server to improve protein engineering...

Area of Science:

• Biochemistry
• Computational Biology
• Protein Engineering

Background:

• Designing proteins with high solubility and low aggregation is vital for biotechnology and medicine.
• Current deep learning methods like ProteinMPNN may not always yield proteins with optimal biophysical properties.

Purpose of the Study:

• To present SPASE (Soluble Protein Analog Selection Engine), an automated webserver for designing proteins with enhanced solubility and reduced aggregation.
• To integrate ProteinMPNN with solubility and aggregation prediction tools for improved protein variant selection.

Main Methods:

• SPASE integrates ProteinMPNN, Protein-Sol, Aggrescan3D, and ESMFold.
• It generates protein variants, predicts solubility and aggregation, models structures, and scores candidates.
• The workflow prioritizes designs based on predicted solubility, aggregation, and folding confidence.

Main Results:

• SPASE computationally enriches for protein analogs with higher predicted solubility.
• The server identifies variants with lower predicted aggregation propensity compared to standard ProteinMPNN output.
• Benchmarking demonstrates SPASE's effectiveness in prioritizing biophysically favorable protein designs.

Conclusions:

• SPASE offers a practical, accessible platform for prioritizing protein engineering candidates.
• Integrated computational approaches are valuable for optimizing multiple biophysical properties in protein design.
• The server facilitates downstream experimental evaluation of designed proteins.