Building molecular model series from heterogeneous CryoEM structures using Gaussian mixture models and deep neural...

Area of Science:

• Structural Biology
• Biophysics
• Computational Biology

Background:

• Cryo-electron microscopy (CryoEM) enables near-atomic resolution structural determination of macromolecules.
• Building accurate molecular models from CryoEM data, especially with conformational heterogeneity, is challenging and labor-intensive.
• Ensuring good stereochemical geometry in these models is crucial for biological interpretation.

Purpose of the Study:

• To present an automated model refinement protocol for CryoEM datasets.
• To address challenges in building high-quality molecular models from heterogeneous CryoEM data.
• To generate molecular models that accurately represent macromolecular dynamics and possess excellent stereochemical geometry.

Main Methods:

• Development of an automated protocol for molecular model refinement.
• Application of the protocol to CryoEM datasets, including those with conformational heterogeneity.
• Generation of multiple molecular models to describe system dynamics.

Main Results:

• The protocol automatically generates series of molecular models.
• The refined models exhibit near-perfect stereochemical geometry scores.
• The generated models effectively describe the dynamics of the macromolecular system.

Conclusions:

• The presented protocol automates and improves the quality of molecular model building from CryoEM data.
• This approach is particularly beneficial for datasets exhibiting conformational heterogeneity.
• The method facilitates a more comprehensive understanding of macromolecular dynamics through improved structural modeling.