Parameterizing elastic network models to capture the dynamics of proteins.

Area of Science:

• Computational Biology
• Biophysics
• Structural Biology

Background:

• Protein dynamics are crucial for function and often studied using coarse-grained normal mode analyses.
• Elastic network (EN) models represent protein geometry as a network of residues to compute fluctuations.
• Current EN models vary in edge selection and parameterization, impacting accuracy.

Purpose of the Study:

• To develop novel tools for constructing more accurate elastic network (EN) models for protein dynamics.
• To improve the geometric representation and parameterization of ENs.
• To enhance the ability of EN models to reproduce experimental and simulation data.

Main Methods:

• Comparison of different geometric models for EN construction, including cutoff-based and Delaunay-based networks.
• Development of an analytical method for parameterizing ENs using flexibility constants per atom.
• Formulation of a non-linear optimization problem for parameterization, accounting for rigid-body and internal motions.
• Validation against experimental B-factors and comparison with molecular dynamics (MD) simulations.

Main Results:

• Delaunay-based ENs provide superior geometric representations compared to cutoff-based ENs.
• The proposed analytical parameterization method yields ENs whose dynamics align well with experimental B-factors.
• ENs parameterized with flexibility constants better mimic MD simulations than those with uniform force constants.
• The refined ENs require fewer normal modes to reproduce functional conformational changes.

Conclusions:

• New tools enhance the construction and parameterization of elastic network models for protein dynamics.
• Delaunay tessellation offers a more accurate geometric basis for ENs.
• Atom-specific flexibility constants improve the predictive power of EN models, reducing computational cost.
• These advancements lead to more reliable and efficient simulations of protein conformational changes.