Computationally designing enzymes with activity levels matching those seen in nature remains a formidable challenge that often requires extensive laboratory optimization. Writing in Nature, Listov et al. now overcome this issue, describing a method that uses atomistic modeling to computationally design highly efficient de novo enzymes.
The authors applied their approach to design a catalyst for Kemp elimination (KE), a non-natural proton abstraction reaction that serves as a model for de novo enzyme design. The workflow uses natural protein backbone fragments to assemble and stabilize backbone variations that are likely to put the resulting enzyme in a catalytically competent constellation. Then geometric matching and Rosetta atomistic calculations are used to position the KE enzyme in each of these structures and to optimize the active site through mutations. Seventy-three designs were selected for experimental testing, of which three showed KE activity. Low-throughput screening further increased their catalytic efficiency. The best-performing design contained more than 140 mutations and an active site constellation different from natural scaffolds.
