Automated design of antibodies, enzymes, and vaccine immunogens

Speaker: Sarel Fleishman
Host: Amelie Stein

Abstract
Many natural proteins are marginally stable or exhibit weak binding or catalytic activity. Conventional methods to improve protein stability and activity, such as in vitro evolution, rely on iterations of genetic mutation and selection. Although these methods are widely and successfully used, they are laborious and may lead to a dead-end where one desired property, such as catalytic rate, can only be improved by trading off another, such as selectivity or stability. To address these fundamental challenges in protein design and engineering, we developed a general strategy that combines phylogenetic sequence analysis with atomistic Rosetta design calculations1. Using this strategy, we developed the PROSS algorithm and web server (http://pross.weizmann.ac.il) for fully automated protein stability design; for instance, we designed the first variants of the human enzyme acetylcholinesterase that are expressible, stable and fully functional in bacterial cells. We then extended the method to design challenging microbial proteins that may serve as vaccine immunogens and applied it to designing a variant of the malaria parasite protein RH5 that can be produced economically in bacterial cells and is stable in the elevated temperatures typical of sub-Saharan Africa where malaria is prevalent. We also developed methods to design the active site of enzymes or binding proteins resulting in orders of magnitude improvement in affinity and catalytic efficiency, including the design of nerve-agent hydrolases exhibiting therapeutically relevant catalytic efficiency. All the above methods design the sequence of the protein given a specific backbone, but large changes in protein activity often depend on substantial changes to the backbone, including insertions and deletions. To address this challenge, we developed a strategy for modular backbone assembly and design and applied it to design accurate new antibodies, high-efficiency enzymes, and ultrahigh specificity protein interaction networks. The automation and generality of this design strategy may enable complete computational design of stable antibodies or enzymes with high affinity, specificity, selectivity, and rate.