Description
Fully automated De-novo design of enzymes with custom catalytic functions still holds as a major challenge in protein design. The recently published method Rfdiffusion has shown great promise for enzyme design because of its ability to scaffold minimal protein motifs that contain active sites extracted from natural enzymes. Here we present an end-to-end pipeline for de novo enzyme design that starts by constructing a library of artificial enzyme motifs from any given geometry of active-site amino acids by placing their inverted rotamers into short secondary structure elements. Next, the pipeline utilizes Rfdiffusion, ProteinMPNN and Rosetta to scaffold the artificial motifs which are subsequently evaluated using the structure prediction methods ESMFold and AlphaFold2. With our pipeline, we have designed novel enzymes that hold substrate binding pockets and exhibit sub-angstrom sidechain RMSD to the input geometry for active site motifs comprised of up to four catalytic amino acids.| Period | 26 Sept 2023 |
|---|---|
| Event title | European Rosettacon: Protein design in the age of artificial intelligence |
| Event type | Conference |
| Location | Leipzig, GermanyShow on map |
| Degree of Recognition | International |