Introduction
A fundamental challenge in immunology is to develop new approaches to decoding the molecular language of immunity for better design and discovery of precision immunodiagnostics and immunotherapeutics. Antibodies are natural diagnostics and therapeutic immune molecules that recognize foreign agents such as viruses, bacteria and vaccines (1). Upon entry, antibodies recognize and neutralize these targets. Almost all vaccine responses are antibody-dependent any antibodies are among the top-grossing drugs for cancer therapy, autoimmunity and many infectious diseases. Specifically, as of yet, the rules that govern antibody-antigen interaction are unclear.
We have recently provided the proof of principle that antibody-antigen is a priori predictable (2–4). However, the analyzed data was too small for a detailed understanding of antibody-antigen interaction. A large symmetric antibody-antigen binding dataset with an equal number of antibodies and potential epitopes is required.
Project background
The ultimate goal of this project is to develop software that can be used to design new therapies for diseases such as infections, auto-immunity, and cancer. This will be a breakthrough in medicine. We have hypothesized the project based on the following rationale:
- Specifically, we will design antibodies, currently the fastest-growing type of new therapeutics.
- "The holy grail" in drug development is software that uses the DNA sequence of, e.g., a viral protein as input and returns the sequence of a therapeutic antibody.
- To develop such software, one must test tens of thousands of antibodies against tens of thousands of proteins to train machine learning algorithms.
- We have unique laboratory biotech techniques and computer algorithms that can be used for this purpose.
In order to generate massive data of binding pairs, we are looking for approaches with the possibility of studying multiple binding pairs simultaneously and the generation of high-throughput data for machine learning. In this study, we will use conventional yeast display technology coupled to flow cytometry-based microsphere affinity proteomics technology (MAP) (5) to screen antibody-antigen binding pairs.
In this project, we aim to establish a robust high-throughput framework for generation of massive data of antibody-antigen binding pairs. The generated data will be analyzed using a variety of bioinformatics tools and used for training machine learning models for antibody-antigen binding prediction.
Project goal
The candidate will work with robotics and proteomics tools to test the binding of tens of thousands of antibodies to tens of thousands of proteins. The master student will be mainly involved in generating antibody and antigen libraries and performing the binding experiments of the expressed libraries using yeast display technology coupled to flow cytometry-based microsphere affinity proteomics technology (MAP). In addition, the student will be able to learn the basics of bioinformatics analysis, including sequencing data preprocessing and computational structural biology
Research environment
The project will be performed at the Computational and Systems Immunology research group (Greiff lab), and the experimental part will be performed at the Proteomic lab (Lund-Johansen lab) located at the Department of Immunology, Rikshospitalet. As a member of the research group, you will be involved in all activities, including weekly scientific meetings, journal clubs and project discussions.
References
- R. Akbar, H. Bashour, P. Rawat, P. A. Robert, E. Smorodina, T.-S. Cotet, K. Flem-Karlsen, R. Frank, B. B. Mehta, M. H. Vu, T. Zengin, J. Gutierrez-Marcos, F. Lund-Johansen, J. T. Andersen, V. Greiff, Progress and challenges for the machine learning-based design of fit-for-purpose monoclonal antibodies. MAbs. 14, 2008790 (2022).
- R. Akbar, P. A. Robert, M. Pavlovi?, J. R. Jeliazkov, A compact vocabulary of paratope-epitope interactions enables predictability of antibody-antigen binding. bioRxiv (2019) (available at https://www.biorxiv.org/content/10.1101/759498v3.abstract).
- R. Akbar, P. A. Robert, C. R. Weber, M. Widrich, R. Frank, M. Pavlovi?, L. Scheffer, M. Chernigovskaya, I. Snapkov, A. Slabodkin, B. B. Mehta, E. Miho, F. Lund-Johansen, J. T. Andersen, S. Hochreiter, I. H. Haff, G. Klambauer, G. K. Sandve, V. Greiff, In silico proof of principle of machine learning-based antibody design at unconstrained scale. bioRxiv (2021), p. 2021.07.08.451480.
- P. A. Robert, R. Akbar, R. Frank, M. Pavlovi?, M. Widrich, I. Snapkov, M. Chernigovskaya, L. Scheffer, A. Slabodkin, B. B. Mehta, M. H. Vu, A. Prósz, K. Abram, A. Olar, E. Miho, D. T. T. Haug, F. Lund-Johansen, S. Hochreiter, I. H. Haff, G. Klambauer, G. K. Sandve, V. Greiff, One billion synthetic 3D-antibody-antigen complexes enable unconstrained machine-learning formalized investigation of antibody specificity prediction, , doi:10.1101/2021.07.06.451258.
- K. Sikorski, A. Mehta, M. Inngjerdingen, F. Thakor, S. Kling, T. Kalina, T. A. Nyman, M. E. Stensland, W. Zhou, G. A. de Souza, L. Holden, J. Stuchly, M. Templin, F. Lund-Johansen, A high-throughput pipeline for validation of antibodies. Nature Methods. 15 (2018), pp. 909–912.