Designing novel solenoid proteins with in silico evolution.

Area of Science:

• Protein engineering
• Computational biology
• Biophysics

Background:

• Solenoid proteins, characterized by tandem repeats, are crucial for various biological functions and are key targets for protein design.
• Machine learning advancements have significantly improved understanding of protein sequence-structure relationships, paving the way for de novo protein design.

Purpose of the Study:

• To develop an in silico evolution platform for de novo design of solenoid proteins.
• To explore the design space of alpha-, beta-, and alpha-beta solenoid backbones.
• To validate computational designs through experimental characterization.

Main Methods:

• Utilized a genetic algorithm coupled with a solenoid discriminator network and AlphaFold2 as an oracle.
• Generated random sequences to design various solenoid backbones.
• Experimentally characterized 41 designed solenoid proteins.

Main Results:

• Successfully designed alpha-solenoid backbones that consistently folded as intended, with one structurally validated design.
• Initial beta-solenoid designs failed, highlighting the challenges in designing beta-strand rich proteins.
• Refined beta-solenoid designs, incorporating terminal capping elements, resulted in two proteins with expected biophysical properties.

Conclusions:

• The developed platform enables fold-specific, hallucination-based de novo protein design without reliance on explicit structural templates.
• The study demonstrates the feasibility of designing novel solenoid proteins computationally and validating them experimentally.
• This approach expands the possibilities for creating proteins with tailored functions.