logo

DeepUMQA is an Ultrafast Shape Recognition (USR)-based model quality assessment method for single protein (monomer) structure and multimeric complexes. We proposed monomer USR to characterize the relationship between local residues and overall topology in monomer structures, and complex USR to characterize the relationship between a residue of one monomer and the topology of other monomers in complex structures. For the monomer structure model, we use model-dependent features, co-evolutionary and template features to represent it, and a enhanced residual neural network based on triangle update and axial attention is proposed to predict the lDDT of each residue. For the complex structure model, we represent it from the overall complex level, intra-monomer level, and inter-monomer level. And the same network as the monomer is used to predict the lDDT of each residue and the accuracy of interface residues, and an equivariant graph neural network is proposed to predict the overall fold accuracy. DeepUMQA (version 3, group name: GuijunLab-RocketX) ranked first in the complex interface residue accuracy estimation of CASP15. DeepUMQA ranked first in the 1-year (2021-12-03 to 2022-11-26) blind test of single protein model quality assessment of CAMEO, and DeepUMQA (version 2) shows state-of-the-art performance in the continuous blind test of CAMEO. (More about DeepUMQA...)


DeepUMQA online webserver
Monomer structure assessment
Input the monomer structure model in PDB format (mandatory, Click for an example input).

Or, upload the monomer structure model file (ends with ".pdb"):

If you want to evaluate multiple models of the same protein and pick the best model (or rank the models), click the "Add model" and "Remove model" buttons below to add or remove structural models.
 
You can also upload a zip file containing all models for the same protein (each model file ends with ".pdb"):

Option:

Do not use templates (check this box if you DO NOT want to use any PDB templates.)

Do not use any homologous sequences (check this box if you DO NOT want to use any homologous sequences.)

Email: (optional, where results will be sent to)

Job name: (optional, your given name to this job)


DeepUMQA News

  • 2022/12/13: DeepUMQA3 (Group name: GuijunLab-RocketX) ranked first in the complex interface residue accuracy estimation of CASP15.
  • 2022/12/01: DeepUMQA updated to DeepUMQA3 for multimeric complexes and interface residue accuracy estimation.
  • 2022/11/26: DeepUMQA ranked first in the 1-year (2021-12-03 to 2022-11-26) blind test of CAMEO-QE.
  • 2022/10/10: The server adds model refinement functioin based on DeepUMQA2.
  • 2022/10/06: DeepUMQA2 ranked first in the 6-month (2022-04-08 to 2022-10-01) blind test of CAMEO-QE.
  • 2022/06/02: DeepUMQA ranked first in the 6-month (2021-12-03 to 2022-05-28) blind test of CAMEO-QE.
  • 2022/05/10: The server adds multiple model quality assessment and ranking functioin.
  • 2022/04/10: DeepUMQA has been updated to DeepUMQA2 by integrating MSA and template features and enhanced network based on triangle update and axial attention.
  • 2022/04/07: DeepUMQA2 ranked first in the 1-month (2022-03-11 to 2022-04-02) blind test of CAMEO-QE.
  • 2021/12/22: DeepUMQA ranked first in the 1-month (2021-11-26 to 2021-12-18) blind test of CAMEO-QE.

Reference

  • Jun Liu, Dong Liu, Guijun Zhang*. DeepUMQA3: a web server for accurate assessment of interface residue accuracy in protein complexes.   Bioinformatics, 2023, 39(10): btad591.
  • Jun Liu, Dong Liu, Guangxing He, Guijun Zhang*. Estimating protein complex model accuracy based on ultrafast shape recognition and deep learning in CASP15.   Proteins: Structure, Function, and Bioinformatics, 2023, 1-10. doi: 10.1002/prot.26564。
  • Jun Liu, Kailong Zhao and Guijun Zhang*. Improved model quality assessment using sequence and structural information by enhanced deep neural networks.   Briefings in Bioinformatics, 2023, 24(1): bbac507.
  • Sai-Sai Guo✝, Jun Liu✝, Xiao-Gen Zhou, Gui-Jun Zhang*. DeepUMQA: Ultrafast Shape Recognition-based Protein Model Quality Assessment using Deep Learning.   Bioinformatics, 2022, 38(7): 1895-1903.