Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

ABodyBuilder3: Improved and scalable antibody structure predictions

Henry Kenlay, Frédéric A. Dreyer, Daniel Cutting, Daniel Nissley, Charlotte M. Deane

TL;DR

Antibody structure prediction is challenged by CDRH3 diversity. ABodyBuilder3 advances the field by combining scalable architecture, ProtT5-based residue embeddings, and a pLDDT-based uncertainty head to deliver state-of-the-art CDR accuracy and practical confidence estimates without ensembles. The approach demonstrates faster training and inference, robust data handling, and effective refinement, with potential to accelerate large-scale therapeutic screening. The work provides a public release of code and weights to support broad adoption.

Abstract

Accurate prediction of antibody structure is a central task in the design and development of monoclonal antibodies, notably to understand both their developability and their binding properties. In this article, we introduce ABodyBuilder3, an improved and scalable antibody structure prediction model based on ImmuneBuilder. We achieve a new state-of-the-art accuracy in the modelling of CDR loops by leveraging language model embeddings, and show how predicted structures can be further improved through careful relaxation strategies. Finally, we incorporate a predicted Local Distance Difference Test into the model output to allow for a more accurate estimation of uncertainties.

ABodyBuilder3: Improved and scalable antibody structure predictions

TL;DR

Antibody structure prediction is challenged by CDRH3 diversity. ABodyBuilder3 advances the field by combining scalable architecture, ProtT5-based residue embeddings, and a pLDDT-based uncertainty head to deliver state-of-the-art CDR accuracy and practical confidence estimates without ensembles. The approach demonstrates faster training and inference, robust data handling, and effective refinement, with potential to accelerate large-scale therapeutic screening. The work provides a public release of code and weights to support broad adoption.

Abstract

Accurate prediction of antibody structure is a central task in the design and development of monoclonal antibodies, notably to understand both their developability and their binding properties. In this article, we introduce ABodyBuilder3, an improved and scalable antibody structure prediction model based on ImmuneBuilder. We achieve a new state-of-the-art accuracy in the modelling of CDR loops by leveraging language model embeddings, and show how predicted structures can be further improved through careful relaxation strategies. Finally, we incorporate a predicted Local Distance Difference Test into the model output to allow for a more accurate estimation of uncertainties.
Paper Structure (7 sections, 3 figures, 3 tables)

This paper contains 7 sections, 3 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Improved structure modelling and evaluation
  3. Language model representation
  4. Uncertainty estimation
  5. Conclusions
  6. Structure refinement
  7. Spearman correlation of uncertainties

Figures (3)

  • Figure 1: Left: Overview of an antibody structure, with the variable region and CDR loops shown. Right: Schematic representation of the ABodyBuilder3 architecture, with 8 sequential and independent update blocks providing the final atomic coordinates and uncertainty predictions from an embedding representation of the variable region sequence.
  • Figure 2: Left: Structure predicted by ABodyBuilder3, with colouring indicating the pLDDT uncertainty estimate. The ground truth (7T0J) is shown in grey. Right: Distribution of CDRH3 RMSD across different bins of the CDRH3 pLDDT score.
  • Figure 3: YASARA2 refinement (x-axis) compared to OpenMM refinement (y-axis).