Proteins are the most complex molecules in nature, and the vast majority of mutations decrease or even eliminate their functions. We found that using evolutionary information, such as from homologous proteins, can tremendously increase the effectiveness of atomistic design calculations, addressing major challenges of protein design and engineering.
We developed a suite of computational design algorithms to address major problems in protein optimisation, including improving stability and protein expressibility, binding affinity and catalytic efficiency, and specificity. Our methods have enabled optimising very challenging enzymes and binding proteins for therapeutics, vaccines, and green chemistry applications.
The most challenging problem in designing proteins with new or substantially improved activities is designing new active-site backbones. We developed a unique strategy that combines backbone fragment from natural homologs and then designs the sequence of the new backbone to maximise stability and activity. Designs with more than 100 mutations demonstrate high catalytic efficiency.
The next frontier for protein design is an automated strategy for designing huge functional repertoires. We are developing methods for design and synthesis of millions of substantially different variants followed by high-throughput screening to characterise the designs, and ML to deepen our understanding of protein design principles. We are applying this strategy to the design of new hydrolytic enzymes, single-domain camelid antibodies, fluorescent proteins, and therapeutic antibodies. These methods will enable rapid and effective discovery and optimisation of biomolecular activities in protein engineering, synthetic biology, and biotherapeutic design.