Modeling cryo-EM structures in alternative states with AlphaFold2-based models and density-guided simulations.

Area of Science:

• Structural Biology
• Computational Biology
• Biophysics

Background:

• Accurate atomic modeling into cryo-electron microscopy (cryo-EM) maps is essential for protein structure determination.
• Modeling proteins with multiple functional states, especially membrane proteins, is challenging due to conformational flexibility and limited template availability.
• Low cryo-EM map resolution hinders de novo model building for alternative protein conformations.

Purpose of the Study:

• To develop and validate a novel computational approach for refining atomic models into cryo-EM maps, particularly for proteins exhibiting conformational transitions.
• To enhance the accuracy of structural modeling for membrane proteins with multiple functional states where traditional methods fall short.

Main Methods:

• Generating multiple initial protein models using stochastic subsampling of AlphaFold2's multiple sequence alignment (MSA) space.
• Clustering the generated models using structure-based k-means to identify distinct conformational states.
• Performing density-guided molecular dynamics (MD) simulations from representative cluster models.
• Selecting the final atomic model based on cryo-EM map fit and overall model quality.

Main Results:

• The proposed refinement approach significantly improved fitting accuracy compared to single starting point methods.
• Demonstrated enhanced accuracy for three pharmacologically relevant membrane proteins (calcitonin receptor-like receptor, L-type amino acid transporter, alanine-serine-cysteine transporter) undergoing conformational changes.
• Successfully facilitated the building of alternative functional states for these membrane proteins.

Conclusions:

• Ensemble construction using generative artificial intelligence (AI) combined with simulation-based refinement is a powerful strategy for modeling alternative protein states.
• This approach is particularly beneficial for understanding the functional dynamics of membrane proteins and other complex biological systems.
• The method offers a robust solution for structure determination challenges posed by conformational heterogeneity in cryo-EM studies.