ModFOLD6

ModFOLD6 estimates model accuracy (EMA) for three-dimensional protein structures to evaluate local and global quality of predicted protein models.

Key Features:

• Hybrid Quasi-Single Model Approach: Integrates scores from both pure-single model methods and quasi-single model methods using a neural network to enhance local quality estimation.
• Local Quality Score Estimation: Combines outputs from three pure-single model methods and three quasi-single model methods to produce detailed per-residue local quality scores.
• Global Score Estimates: Provides three global score options: ModFOLD6_rank optimized for ranking or selecting the best models, ModFOLD6_cor designed to maximize correlations between predicted and observed scores, and ModFOLD6_global offering balanced performance.
• Performance and Validation: Ranked among top EMA methods in CASP12 blind testing and subjected to continuous automatic evaluation in CAMEO, demonstrating improvements over previous versions and other public servers.

Scientific Applications:

• Model Selection: Assists researchers in selecting the most accurate predicted protein models from sets of candidate structures.
• Structural Biology Studies: Provides assessments of structural integrity and local model accuracy for predicted protein conformations.
• Drug Design and Development: Improves confidence in predicted protein structures used in structure-based drug discovery workflows.

Methodology:

ModFOLD6 integrates scores from pure-single and quasi-single model methods and uses a neural network to synthesize these scores into local and global quality estimates.