Persistent Sheaf Laplacian Analysis of Protein Flexibility

Area of Science:

• Computational Biology
• Structural Biology
• Topological Data Analysis

Background:

• Protein flexibility, quantified by B-factors, is vital for protein functions.
• Accurate prediction of protein flexibility is essential for protein design and drug discovery.
• Existing models like the Gaussian network model (GNM) have limitations.

Purpose of the Study:

• To introduce the persistent sheaf Laplacian (PSL) as a novel tool for modeling and analyzing protein flexibility.
• To accurately predict protein B-factors using the PSL model.
• To compare the PSL model's performance against established methods.

Main Methods:

• Representing local protein atom topology and geometry using multiscale harmonic and non-harmonic spectra of PSLs.
• Developing a blind machine learning prediction method incorporating global and local protein features.
• Utilizing a dataset of 364 proteins for B-factor prediction and validation.

Main Results:

• The PSL model effectively captures protein flexibility and predicts B-factors with high accuracy and robustness.
• PSL model achieved a 32% increase in prediction accuracy compared to the classical GNM.
• The blind machine learning approach further validated the PSL model's effectiveness.

Conclusions:

• The persistent sheaf Laplacian is a powerful and accurate tool for predicting protein B-factors.
• PSL offers significant advantages over traditional methods for analyzing protein flexibility.
• This approach holds promise for advancing protein engineering and drug discovery efforts.