QMEANDisCo

QMEANDisCo evaluates the quality of 3D protein structure models by combining consensus distance constraints with single-model QMEAN parameters to produce global and per-residue quality estimates.

Key Features:

• Composite Scoring Function: Integrates the DisCo consensus distance constraint score with single-model QMEAN parameters to generate combined quality estimates.
• Distance Constraint Score (DisCo): Utilizes distance distributions derived from experimentally determined homologous protein structures to inform model scoring.
• Statistical Potentials and Agreement Terms: Employs statistical potentials of mean force and agreement terms from the QMEAN framework as single-model assessment components.
• Adaptive Weighting via Neural Networks: Uses feed-forward neural networks to adaptively weigh contributions from DisCo and QMEAN parameters for each model.
• Performance and Benchmarking: Has been evaluated in benchmarking contexts including continuous assessment in CAMEO and participation in CASP13, demonstrating competitive performance.
• Global and Local Estimates: Produces both global model scores and per-residue local quality estimates.

Scientific Applications:

• Structural model validation: Assessing the reliability of predicted 3D protein structures in the absence of experimental references.
• Drug design: Informing selection of structural models for structure-based drug discovery workflows.
• Functional annotation: Guiding interpretation of predicted structures for inferring protein function and features.
• Protein dynamics and modeling: Providing local quality metrics to support studies of conformational variability and model refinement.

Methodology:

DisCo computes consensus distance constraint scores from distance distributions of homologous experimental structures; QMEAN supplies statistical potentials of mean force and agreement terms; feed-forward neural networks are trained to adaptively weight DisCo and QMEAN contributions to produce global and per-residue composite scores.