Fast de novo antibody structure prediction with atomic accuracy

ABSTRACT

Background: In the field of antibody engineering, an essential task is to design a novel antibody whose paratopes bind to a specific antigen with correct epitopes. Understanding antibody structure and its paratope can facilitate a mechanistic understanding of its function. Therefore, antibody structure prediction from its sequence alone has always been a highly valuable problem for de novo antibody design. AlphaFold2 (AF2), a breakthrough in the field of structural biology, provides a solution to this protein structure prediction problem by learning a deep learning model. However, the computational efficiency and undesirable prediction accuracy on antibody, especially on the complementarity-determining regions limit its applications in de novo antibody design. Method(s) To learn informative representation of antibodies, we trained a deep antibody language model (ALM) on curated sequences from observed antibody space database via a well-designed transformer model. We also developed a novel model named xTrimoABFold++ to predict antibody structure from antibody sequence only based on the pretrained ALM as well as efficient evoformers and structural modules. The model was trained end-to-end on the antibody structures in PDB by minimizing the ensemble loss of domain-specific focal loss on CDR and the frame aligned point loss. Result(s) xTrimoABFold++ outperforms AF2 and OmegaFold, HelixFold-Single with 30+% improvement on RMSD. Also, it is 151 times faster than AF2 and predicts antibody structure in atomic accuracy within 20 seconds. In recently released antibodies, for example, cemiplimab of PD1 (PDB 7WVM) and cross-neutralizing antibody 6D6 of SARS-CoV-2 (PDB 7EAN), the RMSD of xTrimoABFold++ are 0.344 and 0.389 respectively. Conclusion(s) To the best of our knowledge, xTrimoABFold++ achieved the state-of-the-art in antibody structure prediction. Its improvement on both accuracy and efficiency makes it a valuable tool for de novo antibody design, and could make further improvement in immuno-theory.