Bayesian multi-state multi-condition modeling of a protein structure based on X-ray crystallography data

Area of Science:

• Structural Biology
• Biophysics
• Computational Biology

Background:

• Protein atomic structure is determined from crystal X-ray diffraction patterns.
• Crystallography typically yields a single atomic model, but proteins exist in multiple conformational states.
• A single model may not fully represent the protein's dynamic nature and X-ray data.

Purpose of the Study:

• To develop and validate MultiXray, a Bayesian method for multi-state, multi-condition modeling in X-ray crystallography.
• To improve protein structure accuracy by integrating multiple X-ray datasets collected under different experimental conditions.
• To address the challenge of lower data-to-parameter ratios in multi-state modeling.

Main Methods:

• MultiXray utilizes multiple X-ray datasets and a molecular mechanics force field as input.
• It models multiple protein conformational states and their respective weights under each condition.
• Bayesian posterior model density and biased molecular dynamics simulations are employed for model generation and refinement.

Main Results:

• Benchmarking on simulated CypA data showed improved R-free from 0.105 to 0.089 with a second X-ray dataset.
• Application to SARS-CoV-2 Mpro temperature-dependent data improved R-free from 0.253 to 0.237.
• The method successfully integrates diverse experimental data to refine structural models.

Conclusions:

• MultiXray enhances protein structure modeling accuracy by incorporating multiple conformational states and experimental conditions.
• The integration of multiple X-ray datasets significantly improves model quality.
• MultiXray is implemented in open-source Integrative Modeling Platform (IMP) software, compatible with Phenix, facilitating broader application.