Bayesian inference of protein conformational ensembles from limited structural data

Area of Science:

• Structural Biology
• Biophysics
• Computational Biology

Background:

• Proteins feature flexible regions connecting rigid domains, influencing their interactions.
• Understanding protein conformational ensembles in solution is crucial for biological function.
• Small-Angle Scattering (SAS) is a powerful technique for studying protein structures in solution.

Purpose of the Study:

• To develop a robust method for inferring protein conformational ensembles from SAS data.
• To address challenges of limited information content and overfitting in SAS data analysis.
• To improve the characterization of protein flexibility and domain dynamics.

Main Methods:

• A Bayesian statistical framework was developed to infer conformational ensembles.
• Ensembles were generated using all-atom Monte Carlo simulations and a structural library.
• Variational Bayesian inference was used for fast model selection, maximizing model evidence.
• Population weights were determined through complete Bayesian inference.

Main Results:

• The method successfully identifies correct ensembles and recovers the number of members, even with significant noise.
• Model evidence effectively selects appropriate ensemble models.
• Integration with Nuclear Magnetic Resonance (NMR) chemical shifts and structural energies enhances ensemble definition.
• Synergistic use of SAXS, NMR, and energy data improves conformational ensemble accuracy.

Conclusions:

• The developed Bayesian method provides a reliable approach for analyzing protein conformational ensembles using SAS data.
• Combining SAS with NMR and energy calculations significantly refines the characterization of protein dynamics.
• This technique offers a powerful tool for understanding structure-function relationships in flexible proteins.