Adaptive Cartesian and torsional restraints for interactive model rebuilding.

Area of Science:

• Structural biology
• Computational biology
• Biophysics

Background:

• Atomic model building into low-resolution cryo-electron microscopy (cryo-EM) density often uses higher-resolution models as references.
• Over-restraining to reference models can lead to poor fits when experimental data conflicts.
• Current 'top-out' potentials have fixed flattening rates, limiting flexibility.

Purpose of the Study:

• To introduce new restraint potentials with tuneable flattening rates for atomic model refinement.
• To enhance flexibility in encoding confidence for restraints during model building.
• To improve the accuracy of atomic models in low-resolution cryo-EM density.

Main Methods:

• Developed a Cartesian distance restraint based on generalized loss functions.
• Created a periodic torsion restraint using a renormalized von Mises distribution.
• Implemented these potentials as user-adjustable restraints within the ISOLDE software.

Main Results:

• Demonstrated the utility of the new tuneable restraint potentials.
• Showcased their application in real-world atomic model refinement scenarios.
• Provided greater control over the restraint's behavior as model deviation increases.

Conclusions:

• The new tuneable restraint potentials offer enhanced flexibility for atomic model building in low-resolution cryo-EM.
• These methods allow for more nuanced incorporation of prior structural information.
• Improved model accuracy and reliability in structural biology studies can be achieved.