AbAgIntPre

AbAgIntPre predicts antibody–antigen interactions from amino acid sequences to support identification of interacting pairs for therapeutic antibody development.

Key Features:

• Deep Learning Framework: Employs a Siamese-like convolutional neural network (CNN) architecture to analyze antibody and antigen sequence data and predict interactions without requiring 3D structural information.
• Amino Acid Composition Encoding: Represents both antigens and antibodies via an amino acid composition encoding scheme to capture sequence-derived interaction features.
• Performance Metrics: Achieves an Area Under Curve (AUC) of 0.82 on a high-quality generic independent test dataset and shows competitive performance on datasets related to SARS-CoV.
• Complement to Experimental Methods: Provides rapid in silico predictions to complement experimental antibody screening.

Scientific Applications:

• Therapeutic Antibody Development: Supports discovery and optimization of antibodies for therapeutic purposes, including oncology and infectious diseases.
• Rapid Screening: Enables rapid in silico screening of candidate antibody–antigen pairs.
• Research and Drug Design: Provides sequence-based insights into antibody–antigen interactions to inform drug design and studies of immune responses at the molecular level.

Methodology:

A Siamese-like convolutional neural network processes amino acid composition–encoded antibody and antigen sequences to predict interactions from sequence data without requiring 3D structural information.