Three AI methods tackle enzyme design, offering new possibilities for creating custom enzymes with tailored functions. These innovative approaches, including RFdiffusion2, RFdiffusion3, and Riff-Diff, address the challenge of placing catalytic residues with precision in 3D space, a task previously difficult to achieve computationally. The Baker lab's RFdiffusion2, led by Nobel laureate David Baker, revolutionizes enzyme design by starting with an ideal atom cluster, determining catalytic residue placement, and positioning surrounding amino acids and the backbone. This method has shown promise in designing metallohydrolases, enzymes crucial for breaking chemical bonds, with some computer-generated enzymes exhibiting activity, though with lower turnover numbers compared to natural enzymes. The lab further enhances its approach with RFdiffusion3, which designs proteins alongside their interacting molecules, avoiding issues like misfit pockets and unrealistic chemistry. While the success rate is low, these advancements mark significant progress in protein and enzyme design, as acknowledged by organic chemist Kendall Houk. Meanwhile, the Graz team's Riff-Diff focuses on motif-scaffolding, breaking down the enzyme problem into smaller, more manageable parts. By temporarily placing an α-helix in the binding site, Riff-Diff constructs a deeper, more structured pocket, which is then replaced with the intended substrate models. Both teams aim to improve the speed and selectivity of computer-generated enzymes, recognizing the need for a deeper understanding of the catalytic step's importance.
