ATOMDANCE: Kernel-based denoising and choreographic analysis for protein dynamic comparison.

Area of Science:

• Molecular Biology
• Structural Biology
• Computational Biology

Background:

• Protein function arises from nanoscale, nonrandom motions, crucial for molecular interactions.
• Comparative analysis of DNA sequences and protein structures traditionally uses static data.
• Molecular dynamics (MD) simulations offer insights into dynamic protein behavior.

Purpose of the Study:

• To introduce ATOMDANCE, a suite of machine learning tools for denoising comparative MD simulations.
• To identify functional sites and coordinated motions within proteins by analyzing nonrandom dynamics.
• To assess evolutionary and functional shifts in protein dynamics.

Main Methods:

• ATOMDANCE utilizes machine learning for comparative analysis of MD simulations.
• ATOMDANCE-maxDemon4.0 employs Gaussian kernels and maximum mean discrepancy to assess motion differences.
• ATOMDANCE-Choreograph2.0 uses mixed-model ANOVA and graph networks to detect synchronized dynamic shifts.

Main Results:

• ATOMDANCE effectively identifies functional sites in dynamic responses during DNA binding, drug interactions, and virus-host recognition.
• The tool reveals shifts in global and local site coordination during allosteric activation.
• ATOMDANCE enables site-wise testing of neutral versus nonneutral evolution based on dynamic function divergence.

Conclusions:

• ATOMDANCE provides accurate identification of functional protein sites by denoising MD data.
• The suite facilitates understanding of evolutionary dynamics and functional shifts in proteins.
• ATOMDANCE enhances the analysis of complex biological processes through dynamic simulation interpretation.