A novel side-chain orientation dependent potential derived from random-walk reference state for protein fold...

Area of Science:

• Computational biology
• Structural bioinformatics
• Protein structure prediction

Background:

• Accurate potential functions are crucial for protein folding and structure prediction.
• Existing knowledge-based potentials often use inadequate reference states, neglecting protein-specific properties like sequence connectivity and entropic elasticity.

Purpose of the Study:

• To develop novel statistical potential functions for improved protein structure prediction.
• To evaluate the effectiveness of a random-walk chain as a reference state in statistical potentials.

Main Methods:

• Developed a pair-wise, distance-dependent, atomic statistical potential (RW) using an ideal random-walk chain as the reference state.
• Optimized RW on CASP models and benchmarked it on decoy sets.
• Incorporated a side-chain orientation-dependent term into RW (RWplus) to enhance performance.

Main Results:

• RW and RWplus significantly outperform existing pair-wise distance-dependent potentials in selecting native and near-native models.
• These new potentials show higher energy-RMSD and energy-TM-score correlations on decoy sets.
• RW and RWplus demonstrate performance comparable to state-of-the-art scoring functions.

Conclusions:

• A random-walk chain is a suitable reference state, effectively capturing protein sequence connectivity and entropic elasticity.
• RW and RWplus potentials show significant promise for protein structure recognition and folding simulations.
• The RW and RWplus potentials and I-TASSER decoys are publicly available.