Loading...

Binding Affinity Prediction Between Antibody and Antigen using Self-Supervised Learning

Alikhani Ziaratgahi, Mohammad Hassan | 2024

0 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 57473 (19)
  4. University: Sharif University of Technology
  5. Department: Computer Engineering
  6. Advisor(s): Sharifi Zarchi, Ali
  7. Abstract:
  8. In recent years, monoclonal antibodies have gained attention as highly effective drugs for treating diseases, especially cancer. The high binding affinity between an antibody and its corresponding antigen is one of the key factors in triggering an effective immune response. Modeling binding affinity using machine learning is considered a promising and cost-effective computational approach; however, due to the lack of training data, the performance of these models is often poor and limited. In contrast, recent advances in geometric learning have demonstrated that incorporating the three-dimensional geometry of protein structures in the learning process can significantly impact 3D structure-related problems, including binding affinity prediction. In this study, we present a novel approach utilizing self-supervised learning, taking into account the three-dimensional geometry of the antibody-antigen complex structure. This approach models the binding structure of the antibody-antigen complex without requiring training data specifically related to binding affinity and leverages its representation to predict binding affinity. To train the network, we propose self-supervised tasks tailored to the three-dimensional structure of proteins, which lead to learning the physical and chemical properties of the antibody-antigen structure. Then, by designing a regression network as a downstream task, we predict the binding affinity between the antibody and antigen. Finally, by comparing the performance of the proposed approach with previous methods, we found that the use of self-supervised learning can significantly improve the performance of deep learning models. The proposed method showed a 4% improvement in the Pearson correlation coefficient compared to previous works
  9. Keywords:
  10. Binding Affinity ; Self-Supervised Learning ; Cancer Immunotherapy ; Mono Clonl Antibody Immobilization ; Epitope ; Paratope ; Immune System ; Antibody ; Antigen

 Digital Object List

 Bookmark

...see more