GeoAB: Towards Realistic Antibody Design and Reliable Affinity Maturation

Haitao LIN,Lirong Wu,Huang Yufei,Yunfan Liu,Odin Zhang,Yuanqing Zhou,Rui Sun,Stan Z Li

DOI: https://doi.org/10.1101/2024.05.15.594274

2024-09-25

Abstract:Increasing works for antibody design are emerging to generate sequences and structures in Complementarity Determining Regions (CDRs), but problems still exist. We focus on two of them: (i) authenticity of the generated structure and (ii) rationality of the affinity maturation, and propose GeoAB as a solution. In specific, GeoAB-Designer generates CDR structures with realistic internal geometries, composed of a generative geometry initializer (Geo-Initializer) and a position refiner (Geo-Refiner); GeoAB-Optimizer achieves affinity maturation by accurately predicting both the mutation effects and structures of mutant antibodies with the same network architecture as Geo-Refiner. Experiments show that GeoAB achieves state-of-the-art performance in CDR co-design and mutation effect predictions, and fulfills the discussed tasks effectively.

Bioinformatics

What problem does this paper attempt to address?

The paper aims to address two main issues in antibody design:

1. **Authenticity of Generated Structures**: Existing methods often overlook geometric constraints in protein structures, such as peptide planarity and invariant geometric properties (e.g., bond lengths) when designing Complementarity-Determining Regions (CDRs). This results in significant differences between the generated CDR structures and real protein structures. The proposed method, GeoAB-Designer, addresses this issue by generating CDR structures with realistic internal geometric properties.

2. **Rationality of the Affinity Maturation Process**: Affinity maturation refers to the process of enhancing antibody binding affinity by mutating amino acids within the CDR region. Existing methods typically predict the change in binding free energy (ΔΔG) after mutation using trained models and then optimize using the Iterative Target Augmentation (ITA) algorithm. However, this approach has two problems: first, due to the unknown structure of mutants, the input structure required for ΔΔG prediction is usually predicted by other methods, leading to unreliable results; second, inaccurate ΔΔG predictions cannot reliably guide the affinity maturation process. GeoAB-Optimizer addresses these issues by simultaneously predicting mutation effects and post-mutation structures through a dual Geo-Refiner architecture with shared parameters and a joint training strategy.

In summary, the proposed GeoAB framework aims to generate antibody CDR structures that better conform to physicochemical rules and more accurately predict mutation effects, thereby achieving reliable affinity maturation.