proABC-2

proABC-2 predicts antibody paratope residues from heavy- and light-chain sequences using a convolutional neural network to support modeling of antibody–antigen interactions and rational monoclonal antibody design.

Key Features:

• Convolutional Neural Network: employs a CNN architecture that replaces the original proABC random-forest approach to improve pattern recognition within antibody sequences for paratope prediction.
• HADDOCK integration: generates paratope predictions that can be used to define restraints or guide docking simulations in HADDOCK (High Ambiguity Driven protein-protein DOCKing).
• Input: uses antibody heavy- and light-chain sequences as the primary input for prediction.

Scientific Applications:

• Paratope mapping for antibody engineering: identifies paratope residues to support rational design and engineering of monoclonal antibodies.
• Antibody–antigen interaction modeling: informs and improves structural modeling and docking of antibody–antigen complexes when combined with HADDOCK.
• Therapeutic discovery: aids immunotherapy and drug development by characterizing determinants of antigen recognition.

Methodology:

Implements a convolutional neural network that processes input heavy- and light-chain sequences through multiple layers to identify patterns and spatial hierarchies associated with paratope regions; predictions are intended for use with HADDOCK docking simulations.