Scoring predictive models using a reduced representation of proteins: model and energy definition

Area of Science:

• Computational biology
• Structural bioinformatics
• Biophysics

Background:

• Reduced protein models are crucial for studying protein folding.
• Existing models often lack intermediate resolution and robust energy functions for quality assessment.
• The PC2CA model offers a pseudo-covalent structure with two interaction centers per amino acid.

Purpose of the Study:

• To develop a discrete empirical potential for the PC2CA reduced protein model.
• To create an energy function capable of assessing protein model quality.
• To identify native or native-like protein structures within decoy sets.

Main Methods:

• Converted protein structures into a reduced PC2CA form.
• Derived potentials of mean force from distributions of structural parameters (pseudobonds, pseudoangle, pseudodihedrals, distances).
• Incorporated excluded volume effects, finite protein size, and cooperative effects in secondary structures.

Main Results:

• The PC2CA model successfully identified native-like structures (RMSD < 2.0 Å) in decoy sets from the Decoys'R'us database.
• The energy function demonstrated effectiveness in recognizing correct protein folds.
• Performance comparable to scoring potentials with finer granularity was observed.

Conclusions:

• The proposed PC2CA model and energy definition are effective for protein model quality assessment.
• The model shows potential for rapid exploration of protein conformational space.
• Results align with findings from the CASP7 quality assessment experiment.