Deep Learning-based structure modelling illuminates structure and function in uncharted regions of β-solenoid fold...

Area of Science:

• Structural biology
• Computational biology
• Protein science

Background:

• Repeat proteins are ubiquitous, featuring diverse functions and structural motifs.
• Beta-solenoids, a class of repeat proteins, possess unique structural characteristics like handedness and twist.
• Many beta-solenoid proteins, crucial in virulence factors and functional amyloids, lack experimental structures due to modeling and crystallization challenges.

Purpose of the Study:

• To discover novel beta-solenoid protein structures using deep learning.
• To identify structural neighbors and infer potential functions of predicted beta-solenoids.
• To explore the structural diversity and functional implications of beta-solenoid folds.

Main Methods:

• Application of various deep learning-based structure-modeling techniques.
• Utilizing structural database searches to find homologous proteins.
• Relating predicted structures to potential biological functions.

Main Results:

• Discovery of novel eukaryotic and prokaryotic adhesins with beta-solenoid folds.
• Identification of exceptionally long, flat beta-solenoid structures potentially forming mucin tandem repeats.
• Characterization of unprecedentedly small beta-solenoid structures and a novel FapC Greek key beta-solenoid fold.

Conclusions:

• Deep learning effectively predicts diverse beta-solenoid structures and their functions.
• Confirms the link between beta-solenoid folds and adhesin functionality.
• Highlights potential roles in mucin structures and Pseudomonas functional amyloid fibers.