Lattice protein folding with two and four-body statistical potentials

Area of Science:

• Computational Biology
• Structural Bioinformatics
• Biophysics

Background:

• Protein folding is a complex process often described by statistical potentials.
• Multibody potentials generalize two-body potentials using residue cluster probability distributions and the Boltzmann condition.

Purpose of the Study:

• To compare a novel four-body statistical potential with the established Miyazawa-Jernigan (MJ) potential for protein structure prediction.
• To evaluate the effectiveness of a four-body potential derived from Delaunay tessellation of protein structures.

Main Methods:

• A lattice chain growth algorithm was employed for protein structure prediction.
• A four-body statistical potential was used as a discriminatory function for conformational ensembles generated by the MJ potential.
• The study analyzed 22 proteins ranging from 30 to 76 residues in length.

Main Results:

• The four-body potential achieved results comparable to the two-body MJ potential, with an average coordinate root-mean-square deviation (cRMSD) of 8 Å for all-alpha proteins.
• Performance was somewhat lower for protein classes other than all-alpha proteins.
• Superpositions of predicted and native structures demonstrated rough overall agreement for both potentials.

Conclusions:

• The four-body statistical potential shows promise for protein structure prediction, yielding results comparable to the widely used MJ potential.
• Further improvements may be achieved by formulating the four-body potential with larger datasets and direct generation of conformational ensembles using multibody potentials.
• This study highlights the potential of multibody potentials in understanding protein folding and structure prediction.