Antibody binding is the basis of antibody-based therapy in cancer and autoimmunity. Antibodies are also of incredible importance in vaccine-based immune protection. Currently, antibody-drug development is slow because it is mostly based on experimental research. In this proposal, we aim to accelerate antibody research by developing novel computational approaches to antibody design.
Our approach to developing novel methods for antibody design is two-fold: we will investigate new biotechnological methods for screening antibody binding at high-throughput. We will use the newly generated data to feed into novel machine learning methods for antibody binding prediction. Finally, we will demonstrate experimentally the usability of our technology for computational antibody design.
Our proposed technology platform unlocks the possibility to design antibody binding, reducing in the long term the time and cost for therapeutics design and increasing the number of druggable targets.
Background: Antibody binding is a fundamental basis of medical treatment and diagnosis. To accelerate the design of antibody therapeutics, we need to address the challenge of in silico antibody (Ab) design using an interdisciplinary approach.
Research question: Currently, the performance of in silico antibody design tools is poor, requiring sophisticated approaches in experimental and in silico dataset and machine learning design.
Approach: To address the suboptimal performance of current antibody design methods, we will develop experimental methods that screen antigen binding of large numbers of different Abs. We will then use these large-scale antibody binding data to develop novel machine learning approaches for designing in silico antigen-specific antibody binding. Structural information on antibody sequences will be obtained using recent advances in computational structural biology. Machine learning approaches will take advantage of both sequence and structure information. Finally, regarding the feasibility of in silico antibody design, we will express the designed antibodies and test them for antigen binding. We will also perform simulations to benchmark machine-learning antibody design approaches.
Long-term impact: Our proposed interdisciplinary technology approach unlocks the possibility of designing Ag-specific antibodies for applied immunobiotechnology, reducing the time and cost of antibody therapeutics design in the long term.