ProQ2

ProQ2 predicts the quality of protein structural models using a neural network-based single-model approach to estimate local and global accuracy.

Key Features:

• Single-Model Methodology: Operates as a single-model predictor rather than relying on ensemble or consensus methods.
• Local and Global Accuracy Estimation: Provides per-residue (local) and overall (global) quality estimates for protein models.
• Integration with Rosetta Suite: Can be used within the Rosetta modeling suite and contributes machine-learned scoring functions to conformational sampling.
• Batch Processing Capability: Supports large local batch runs for processing multiple models simultaneously.
• Benchmark Performance: Demonstrated superior performance in benchmark evaluations including CASP11 and CAMEO-QE among single-model quality estimators.

Scientific Applications:

• Model Selection: Identifies higher-quality protein models from sets of predicted structures.
• Structure Prediction Improvement: Guides refinement by indicating regions of low and high accuracy in models.
• Conformational Sampling: Informs sampling and scoring of alternative conformations via machine-learned scoring functions.

Methodology:

ProQ2 employs a neural network-based single-model approach and machine-learned scoring functions and can be integrated with the Rosetta modeling suite.