DLAB

DLAB applies deep learning to structure-based virtual screening of antibodies to predict antibody–antigen interactions and improve docking pose ranking for antibody therapeutic discovery.

Key Features:

• Structure-Based Deep Learning Framework: Employs structure-based deep learning to predict antibody–antigen interactions, including for antigens with no previously identified binders.
• Virtual Screening Capability: Performs virtual screening of candidate antibodies against specified antigen targets to computationally prioritize potential binders.
• Improved Antibody–Antigen Docking and Pose Ranking: Enhances docking by generating and more accurately ranking binding poses to prioritize likely true binders.
• Identification of Binding Antibodies: Demonstrated ability in case studies to identify binding antibodies against specific antigens.

Scientific Applications:

• Antibody drug discovery and optimization: Enables structure-based prioritization of antibody candidates to accelerate discovery and optimization of therapeutic antibodies.
• Screening for novel antigen binders: Supports virtual identification of candidate binders for antigens lacking known antibodies.

Methodology:

Deep learning models are trained on structural data of known antibody–antigen interactions and applied to predict binding affinities and binding poses for novel antigen targets, including cases with no known binders.