FRST

FRST evaluates predicted protein structures by scoring models with combined knowledge-based potentials to discriminate native-like structures from incorrect decoys.

Key Features:

• Knowledge-Based Potentials: Integrates four knowledge-based potentials—pairwise interactions, solvation effects, hydrogen bonding, and torsion angles—to capture complementary features of native protein structures.
• Orthogonal Information Sources: Leverages the largely orthogonal nature of these potentials so diverse structural characteristics are considered independently.
• Torsion Angle Potential Correlation: Employs a torsion angle potential that demonstrates a strong correlation with overall model quality for distinguishing native-like structures.
• Linear Weighting Function: Combines the individual potentials through a linear weighting function to construct an energy function that discriminates high-quality from suboptimal models.
• Per-Residue and Global Scoring: Computes energy profiles that enable assessment of both overall model quality and per-residue structural integrity.

Scientific Applications:

• Comparative Modeling: Ranks and selects the best model from multiple accurate candidate models generated for a target protein.
• Fold Recognition and Novel Fold Targets: Identifies native-like structures among numerous incorrect models in fold-recognition and novel-fold prediction scenarios.

Methodology:

Computes an energy profile by combining four knowledge-based potentials (pairwise interactions, solvation, hydrogen bonding, torsion angles) via a linear weighting function and evaluates both overall and per-residue structural integrity; performance has been assessed on benchmarking sets including blind tests such as CAFASP-4 MQAP where correct templates were distinguished from decoys.