Path Similarity Analysis: A Method for Quantifying Macromolecular Pathways.

Area of Science:

• Computational Biology
• Biophysics
• Structural Biology

Background:

• Proteins undergo large conformational changes essential for their function.
• Computational methods aim to simulate these macromolecular transition paths.
• Comparing different simulation paths is challenging due to their high-dimensional nature.

Purpose of the Study:

• Introduce a novel method, Path Similarity Analysis (PSA), for quantitative comparison of molecular transition paths.
• Enable assessment of various computational algorithms for sampling protein conformational changes.
• Identify atomic-scale determinants responsible for differences between simulated pathways.

Main Methods:

• Utilized Hausdorff or Fréchet metrics from computational geometry to compare piecewise-linear curves in 3N-dimensional space.
• Avoided dimensionality reduction, preserving full trajectory information.
• Applied PSA to analyze conformational transitions of adenylate kinase (AdK) and diphtheria toxin.

Main Results:

• PSA effectively quantifies similarity between arbitrary molecular paths, distinguishing between algorithms.
• Clustering revealed that paths generated by the same method are more similar than those from different methods.
• Identified specific molecular features, like salt bridges, as determinants of pathway differences between methods (e.g., DIMS MD vs. FRODA).

Conclusions:

• PSA provides a robust framework for comparing and validating protein transition path sampling methods.
• The method enhances the analysis of conformational transitions by revealing atomic-scale structural differences.
• PSA offers a new approach to understanding and analyzing complex molecular dynamics.