Pushing the boundaries of molecular replacement with maximum likelihood

Area of Science:

• Crystallography
• Structural Biology
• Computational Chemistry

Background:

• Molecular replacement (MR) is a powerful technique for determining protein structures.
• Traditional MR methods struggle with complex unit cells and poor models.
• Likelihood-based methods have shown promise in other crystallographic applications.

Purpose of the Study:

• To develop and evaluate a new likelihood function for molecular replacement.
• To improve the performance of MR, particularly for challenging crystallographic problems.
• To enhance the accuracy and sensitivity of MR targets.

Main Methods:

• Developed a novel likelihood function suitable for rotation searches with unknown relative phases.
• Incorporated correlations between sequence identity and coordinate error for model quality calibration.
• Implemented a statistically valid method for combining multiple molecular models using multivariate complex normal distribution.
• Tested the new methods in the molecular replacement program Beast.

Main Results:

• The new likelihood functions are more sensitive and accurate than previous targets.
• The developed method significantly improves the success rate of molecular replacement.
• The multiple-model likelihood function demonstrates a dramatic positive impact on MR success.

Conclusions:

• Likelihood-based approaches offer substantial performance improvements for molecular replacement.
• The new likelihood function effectively addresses challenges in rotation searches.
• The Beast program with its enhanced likelihood functions represents a significant advancement in structural determination.