GOAP

GOAP scores protein structures using a generalized orientation- and distance-dependent all-atom statistical potential to discriminate native folds from decoys.

Key Features:

• All-atom, plane-based representation: Considers heavy atoms and the relative orientation of planes associated with each interacting heavy-atom pair.
• Orientation and distance dependency: Integrates both orientation and distance dependencies in the potential rather than using representative atoms or block-based approximations.
• Decomposition into components: Separates the potential into distance- and angle-dependent contributions for precise modeling of interactions.
• DFIRE reference state: Uses the DFIRE (Distance-scaled Finite Ideal Gas Reference) state for the distance-dependent component.

Scientific Applications:

• Protein structure prediction and model scoring: Ranks candidate protein models to improve selection of native-like conformations.
• Decoy discrimination and benchmarking: In evaluations on 11 decoy sets containing 278 targets, GOAP identified 226 native structures as top-scoring, outperforming DFIRE (127) and showing approximately 15% higher recognition accuracy than OPUS-PSP while RWplus performed poorly under the same tests.

Methodology:

Derives a knowledge-based statistical potential from observed atomic interactions using an all-atom representation, decomposes the potential into distance- and angle-dependent contributions, employs the DFIRE reference state for the distance term, and encodes orientation by analyzing relative orientations of planes associated with heavy atoms; applied to structures including those from homology modeling and ab initio methods like ROSETTA.