Pipeline for inferring protein function from dynamics using coarse-grained molecular mechanics forcefield.

Area of Science:

• Structural Biology
• Computational Biology
• Biophysics

Background:

• Protein dynamics are crucial for function, but molecular dynamics (MD) simulations are underutilized for inferring molecular function.
• Genomics projects often yield novel proteins with limited evolutionary information, necessitating new functional inference methods.

Purpose of the Study:

• To introduce DynFunc, a publicly accessible web service for inferring protein function from coarse-grained (CG) MD trajectories.
• To provide a novel method that analyzes protein flexibility and dynamics to predict molecular function, independent of evolutionary information.

Main Methods:

• Utilized coarse-grained (CG) molecular dynamics (MD) simulations to generate protein trajectory data (≥1 µs).
• Developed a novel approach combining residue fluctuation-graph analysis and auto-correlation vectors from MD trajectories.
• Implemented the method as the DynFunc web server, offering a custom coarse-grained molecular mechanics (CGMM) forcefield.

Main Results:

• Validated the method on diverse protein datasets with sequence identities as low as 3%, achieving high function-recall rates.
• DynFunc successfully identifies flexible protein regions correlated with putative molecular functions.
• The approach demonstrated effectiveness for novel and moonlighting proteins lacking evolutionary data.

Conclusions:

• DynFunc offers a unique, evolution-independent method for protein function inference using MD simulation data.
• The web service is a valuable tool for structural biologists studying novel proteins and moonlighting functions.
• This approach expands the utility of MD simulations for understanding protein function, particularly in the context of genomics.