RFDiffusion: Accurate Protein Design

RosettaCommons

RFdiffusion is a deep learning framework, that refines the RoseTTAFold structure prediction network. This innovative model excels in diverse protein design challenges, including de novo binding, higher-order symmetry, and enzyme active site scaffolding. RFdiffusion's success lies in its ability to generate complex, functional proteins from basic molecular specifications, showcasing its versatility through experimental validations of hundreds of new designs. This marks a significant advancement in protein design using deep learning, overcoming previous limitations in modeling protein backbone geometry and sequence-structure relationships. Furthermore, RFDiffusion outputs can be used in ProteinMPNN to refine these backbones and design sequences that fold correctly and exhibit the desired functionalities. This combined approach allows the creation of new proteins with specific shapes and functions tailored to various needs.

Example parameter inputs:

  • Unconditional Protein Generation:

               Contig Map Input: 150-200

               Hotspot Points Input: None

               Symmetry Options: NaN

  • Motif Scaffolding
               Contig Map Input: 10-40/A163-181/10-40

               Hotspot Points Input: None

               Symmetry Options: NaN

  • Binder Design

               Contig Map Input: A1-150/0 70-100

               Hotspot Points Input: A59,A83,A91

               Symmetry Options: NaN

  • Symmetric Oligomers Generation

               Contig Map Input: 360-360

               Hotspot Points Input: None

               Symmetry Options: Tetrahedral

  • Fold Conditioning - PPI

               Contig Map Input: None

               Hotspot Points Input: A59,A83,A91

               Symmetry Options: NaN

Check the examples page for more information!

Technology: Diffusion Model

Limitations: Some metrics are kept as default. You can check here for more details.
Citation:
Broadly applicable and accurate protein design by integrating structure prediction networks and diffusion generative models Joseph L. Watson, David Juergens, Nathaniel R. Bennett, Brian L. Trippe, Jason Yim, Helen E. Eisenach, Woody Ahern, Andrew J. Borst, Robert J. Ragotte, Lukas F. Milles, Basile I. M. Wicky, Nikita Hanikel, Samuel J. Pellock, Alexis Courbet, William Sheffler, Jue Wang, Preetham Venkatesh, Isaac Sappington, Susana Vázquez Torres, Anna Lauko, Valentin De Bortoli, Emile Mathieu, Regina Barzilay, Tommi S. Jaakkola, Frank DiMaio, Minkyung Baek, David Baker bioRxiv 2022.12.09.519842; doi: https://doi.org/10.1101/2022.12.09.519842
Released: Nov-20-2023
v0.1
Structural Bioinformatics
3D Protein Structure Prediction
Diffusion
pdb
166
20
share
Example Results
View Source
Previous Job Parameters
Your previous job parameters will show up here
so you can keep track of your jobs

Upload PDB File

No files selected!
OR
Use Demo Data
Your file can be in the following formats:
pdb
• The Protein Data Bank (PDB) data format is a standard file format used to store information about the three-dimensional structures of biological macromolecules.

Set Parameters

Unconditional Protein Generation
NaN