AbDev

AbDev predicts 12 biophysical properties of monoclonal antibodies from variable region sequences using deep learning-based surface descriptors and machine learning models.

Key Features:

• DeepSP Integration: Incorporates a deep learning model that directly predicts dynamical and structural properties of spatial aggregation propensity and spatial charge map in different antibody regions from sequences.
• Machine Learning Prediction: Employs machine learning models trained on molecular dynamics simulation data to predict 12 biophysical properties.
• Sequence-Based Input: Accepts variable region sequences for prediction of antibody developability properties.

Scientific Applications:

• Aggregation Propensity Prediction: Predicts the tendency of monoclonal antibodies to aggregate, a critical factor in therapeutic development.
• Solubility Prediction: Assesses antibody solubility to optimize developability as drug candidates.
• Viscosity Prediction: Evaluates viscosity properties relevant to formulation and large-scale production of therapeutic antibodies.
• Developability Assessment: Supports optimization of monoclonal antibodies during research and development to ensure success as therapeutic drug candidates.

Methodology:

AbDev utilizes full-length antibody molecular dynamics simulations combined with machine learning techniques, and integrates the DeepSP deep learning model for predicting spatial aggregation propensity and spatial charge map from sequences.