Assessing protein conformational sampling methods based on bivariate lag-distributions of backbone angles

Area of Science:

• Computational Biology
• Structural Bioinformatics
• Biophysics

Background:

• Protein structure prediction remains a significant challenge in computational biology.
• Angular-sampling methods are increasingly used to model protein conformational space.
• Existing methods often use parametric models for sequential dependencies between protein chain angles.

Purpose of the Study:

• To review and assess angular-sampling-based methods for protein structure prediction.
• To determine optimal distribution types and mixture model components for protein angles.
• To investigate the order of local sequence-structure dependency for prediction.

Main Methods:

• Assessment of bivariate lag-distributions of dihedral/planar angles.
• Application of Lag singular value decomposition (LagSVD) for nonparametric analysis of angle distributions across lags.
• Development of graphical tools and numerical measurements for performance evaluation.

Main Results:

• Comparative analysis of different model fits for protein angle distributions.
• Identification of key information across lags using LagSVD.
• Development of a web-tool for visualizing protein structure prediction dynamics.

Conclusions:

• Provides a framework for evaluating and comparing angular-sampling methods.
• Offers insights into modeling protein dihedral/planar angles for improved prediction.
• Facilitates the development of more accurate protein structure prediction tools.