Protein structure prediction by trRosetta

Introduction

trRosetta is an algorithm for fast and accurate protein structure prediction. It builds the protein structure based on direct energy minimizations with a restrained Rosetta. The restraints include inter-residue distance and orientation distributions, predicted by a deep neural network. Homologous templates are included in the network prediction to improve the accuracy further. Based on trRosetta, our group (Yang-Server) won CASP15 and CASP16 for protein structure prediction. Read more about trRosetta...

The major results returned include (click here for an example):

Five predicted 3D structure models with estimated global and local accuracy
Predicted inter-residue contacts/distances/orientations
The top PDB templates (when detected) and the multiple sequence alignment

Submit

Provide the protein data (mandatory)

Input a protein sequence (Click for an example input) or a multiple sequence alignment (MSA) below.

Or upload the protein sequence/MSA file:

Input type: query sequence query MSA (Click for explanation)

Other information (optional)

Email: (Optional, where the results will be sent to)

Target name: (Optional, your given name to this target)

Do not use templates (check this box if you DO NOT want to use any PDB templates; the library was updated on Mar 13, 2022. Check here for more information)

Run trRosettaX-Single (check this box for single-sequence folding, e.g., no homologous sequences and templates will be used.)

Keep my results private (check this box if you want to keep your job private. A key will be assigned for you to access the results)

News

12/01/2024,

Our group won CASP16! Yang-Server was ranked at the 1st among 110 groups in the Protein Domains.
Yang-Multimer was ranked at the 2nd among 82 groups in the Protein multimers.

12/10/2022, Breaking news The trRosetta server (as 'Yang-Server') was ranked on the top (among 132 participating groups) in the CASP15 experiment.

11/07/2022, The paper for trRosettaX-Single was accepted for publication by Nature Computational Science. The source codes can be downloaded here.

08/26/2022, The trRosetta server has predicted the structure models for >100,000 proteins.

07/26/2022, Predicted local accuracy was included for each model. The results page was redesigned to show local accuracy.

05/03/2022, The network was updated with an improved accuracy by ~10%.

01/06/2022, The maximum length of protein sequence was increased from 1000 to 1500.

12/23/2021, Single-sequence folding with trRosettaX-Single was included. It works well for orphand proteins and designed proteins.
( >> read more ... ).

References

Du et al, The trRosetta server for fast and accurate protein structure prediction, Nature Protocols, 16: 5634-5651 (2021). (PDF)

Wang et al, Single-sequence protein structure prediction using supervised transformer protein language models, Nature Computational Science, 2: 804-814 (2022). (PDF)

Xiang et al, Generating dynamic structures through physics-based sampling of predicted inter-residue geometries, Advanced Science, 13: e18469 (2026). (PDF)