Computational Design of Novel Enzymes Guided by Evolutionary Data

The ability to computationally design efficient, specific enzymes is a rigorous test of our understanding of the principles of catalysis and molecular recognition.

Successful designs have to date shown several limitations: they only targeted simple reactions, involving two to three catalytic residues with low efficiencies and selectivities, and impaired stability. We developed a new algorithm using Rosetta to combine compatible backbone fragments from natural enzymes of the same enzyme superfamily to generate novel conformations. The designs’ sequences are then optimized, guided by sequence conservation data to improve stability and expressibility. We used the algorithm to design novel TIM barrel fold enzymes belonging to the GH10 family capable of hydrolyzing xylan, an abundant plant polysaccharide, with Kcat/Km values similar to those of natural xylanases. The designed enzyme conformations differ from one another and from any other known natural xylanase conformations and have different substrate specificities.

The algorithm is completely automated and can be applied to other enzymes of modular fold to efficiently and broadly explore the potential selectivities of the superfamily.