DISTEVAL

DISTEVAL evaluates predicted protein inter-residue distances and provides quantitative and qualitative assessments to support protein structure prediction.

Key Features:

• Supported prediction types: Handles real-valued distances and binned distograms as input representations.
• Input flexibility: Accepts predicted contacts, predicted distances, and an optional true structure for comparison.
• Visualizations: Generates heatmaps, chord diagrams, and 3D models to represent predicted distance relationships and residue interactions.
• Quantitative metrics: Computes Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), and contact precision for evaluation.

Scientific Applications:

• Evaluation of protein structure predictions: Provides quantitative and visual assessments of inter-residue distance predictions to inform model accuracy.
• Refinement of predictive models: Supplies metrics and visual outputs that support researchers in refining distance- and contact-prediction methods.
• Interpretation of structural details: Facilitates analysis of residue spatial relationships relevant to biological functions and interactions.

Methodology:

Accepts predicted contacts, real-valued distances, or binned distograms and an optional true structure; computes MAE, RMSE, and contact precision; and generates heatmaps, chord diagrams, and 3D models.