RDb2C2

RDb2C2: ResNet-based β-β residue pairing predictor

RDb2C2 predicts residue-residue β-β pairings within β strands from amino acid sequences using a residual neural network (ResNet) architecture to improve prediction accuracy over the multi-stage random forest framework implemented in RDb2C.

Key Features:

• Residual Neural Network Architecture: Implements a ResNet model to enhance prediction of residue-residue interactions in β sheets.
• Improved Benchmark Performance: Achieves >10 percentage point increase in F1-score over previous methods, with ~72% accuracy on BetaSheet916 and ~73% accuracy on BetaSheet1452 datasets.

Scientific Applications:

• β-Protein Structural Modeling: Improves modeling of mainly β proteins by providing accurate β-β pairing predictions for structural biology and protein folding studies.

Methodology:

Applies a residual neural network (ResNet) to amino acid sequence data to predict β-β residue pairings, replacing the prior multi-stage random forest framework used in RDb2C and enabling improved residue-residue interaction prediction performance.