Statistical machine learning for comparative protein dynamics with the DROIDS/maxDemon software pipeline

Area of Science:

• Biophysics
• Computational Biology
• Structural Biology

Background:

• Traditional protein analysis often overlooks crucial dynamic and functional aspects.
• Understanding protein dynamics is key to comprehending biological function and disease mechanisms.

Purpose of the Study:

• To develop and present a computational pipeline with a graphical user interface for analyzing protein dynamics.
• To enable biophysical simulation and statistical comparison of two-state protein dynamics.
• To identify and compare functionally conserved dynamic motions in protein variants using machine learning.

Main Methods:

• Development of a computing pipeline integrated with molecular visualization tools.
• Biophysical simulation of two-state protein dynamics (unbound vs. complex states).
• Application of multi-agent machine learning classifiers to identify conserved dynamic motions.

Main Results:

• The pipeline facilitates the comparison of protein dynamics beyond static sequence or structure alignments.
• Identification of functionally conserved dynamic motions across different protein variants.
• Enables statistical comparison of dynamic differences in genetic or drug-class variants.

Conclusions:

• The developed tool enhances the analysis of protein function by incorporating dynamic information.
• Machine learning classifiers effectively identify conserved dynamic motions relevant to protein function.
• This approach provides a novel method for comparing functional dynamics in protein variants.