Unlock Antibody Design Potential: Paragraph Integration on Vecura

What is Paragraph?

Paragraph is an advanced computational model designed to predict antibody paratope (antigen-binding site) residues directly from 3D structures. Developed by the Oxford Protein Informatics Group (OPIG), it utilizes an Equivariant Graph Neural Network (EGNN) to process IMGT-numbered antibody Fv region structures. By treating the antibody as a graph of C-alpha atoms, the model assigns a probability score to each residue, indicating its likelihood of being involved in antigen binding.

It helps users pinpoint critical binding residues without requiring computationally expensive simulations. It is especially useful for antibody engineering tasks such as affinity maturation, humanization, and general developability assessment.

What can users do with Paragraph on Vecura?

With Paragraph on Vecura, users can:

• Predict paratope membership for every residue in the antibody variable domain.
• Process paired (heavy and light chain), heavy-only, or light-only Fv structures seamlessly.
• Generate easily filterable CSV tables containing per-residue probability scores.
• Facilitate high-throughput screening of antibody libraries due to the model's rapid CPU-based inference.

What the output means

The output provides a CSV table containing per-residue sigmoid probability scores (ranging from 0 to 1). A threshold of 0.734 is recommended to classify residues as part of the paratope.

This output should be used to support scientific decision making. It does not replace experimental validation.

Why this matters

In the field of computational immunology, identifying the exact residues that constitute the paratope is a fundamental challenge. Traditional experimental methods can be resource-intensive, while conventional computational tools often lack the necessary speed or structural sensitivity required for large-scale library screening. Paragraph bridges this gap by leveraging geometric deep learning to ensure predictions remain consistent regardless of protein orientation.

By enabling rapid and accurate paratope prediction, Paragraph allows researchers to prioritize mutation hotspots and focus their wet-lab efforts on the most promising antibody candidates. This significantly streamlines the R&D pipeline for therapeutic antibody discovery.

• Developed by: Oxford Protein Informatics Group (OPIG)
• Source: Official GitHub repository
• Reference: Chinery et al., Bioinformatics 2023