Roche ML for Antibody Design

In a new two-year collaboration with Roche, the group will be supporting a Roche Postdoctoral Fellowship in the development of antibodies. We will make use of Roche’s existing antibody data to develop machine-learning models generating sequences of antibodies with desired biophysical properties. Our aims are to:

1. Develop machine learning models that ingest in-house antibody engineering data on antibodies to generate novel sequences with increased likelihood of desired biophysical properties,
2. Iteratively apply methods to antibody datasets,
3. Collaborate with experimental teams to select and test model-generated candidate sequences in the wet lab,
4. Embed machine learning models in discovery processes.