rrQNet

rrQNet estimates the quality of predicted protein contact maps using deep learning applied directly to two-dimensional residue–residue contact matrices.

Key Features:

• Evolutionary Reconciliation Network: Utilizes cascaded deep neural network modules to encode evolutionary context and perform reconciliation between predicted contacts and evolutionary signals.
• Residue-Pair Level Scoring: Generates quality scores for individual residue pairs and aggregates them to produce overall contact map quality estimates, enabling multi-resolution assessment.

Scientific Applications:

• Protein Structure Prediction Assessment: Evaluates and ranks predicted contact maps to support validation and refinement of computational protein structure models.

Methodology:

rrQNet applies a deep neural network architecture trained on diverse contact predictors to assess consistency between predicted residue–residue contacts and evolutionary information, producing quantitative quality scores for contact maps.