Vaccines are among the most significant achievements in biomedical science, leading to the eradication of smallpox and targeting infectious agents like polio, measles, and rubella for elimination. Despite these successes, creating vaccines for some critical infectious diseases remains challenging. Our research focuses on eliciting antibody-mediated immune responses with vaccines. While monoclonal antibodies (mAbs) can offer broad protection against diseases like HIV-1, pandemic influenza, and Ebola virus, creating vaccines that induce similar antibody properties has been elusive. Accordingly, a major goal of our research is to devise protein-engineering strategies to enable immunofocusing – the creation of vaccines capable of eliciting an antibody response against a targeted epitope.
Additionally, we complement our experimental efforts with machine learning algorithms to reconstruct protein evolution landscapes, predicting evolutionary velocity. This approach aids in guiding artificial evolution and enhancing antibody affinity maturation using protein sequence data alone. While our work established the effectiveness of language models to guide evolution using sequence alone, a protein's function is inherently linked to its structure. By incorporating structural information with an inverse-folding informed language model, we generalize the approach to protein complexes, and demonstrate substantial gains in predictive capabilities, enabling efficient antibody and protein engineering. These results lay the groundwork for more potent and resilient vaccine and therapeutic design.