Dozens of antibodies are in routine clinical use for treating life-threatening diseases. The conventional route to antibody discovery and optimization relies on animal immunization and optimisation of improved variants. These methods are powerful, and yet, they are costly, time consuming, and ultimately rely on trial-and-error. The reason we still rely on inefficient processes despite their low scalability is a lack of control over biomolecular structure and function. The AutoCAb project is dedicated to developing a new computational strategy for designing effective antibodies while increasing our understanding and control over biomolecular recognition, which is one of the most fundamental aspects of protein activity. The AutoCAb project develops cutting-edge computational and experimental methods that allow designing, for the first time, millions of antibodies or other functional proteins, and synthesizing all of these antibodies accurately and economically. Finally, selection experiments followed by next-generation sequencing allow us to monitor which designs bind their targets and to use advanced machine-learning methods to infer rules that would improve the chances of success in the next round of experiments. The major goals of AutoCAb have been achieved including several papers and patent filings that describe the envisioned approaches and their application to antibodies, enzymes, and fluorescent proteins. The methods developed in AutoCAb enable, for the first time, the design of large and very effective libraries of functional proteins. These libraries comprise protein variants that exhibit superior stability and activity to the natural (or engineered) starting point protein and enable efficient and effective optimisation of biomedically and biotechnologicallly important proteins, such as enzymes and antibodies.
