FINDSITE-metal

THIS WEBSERVICE IS TEMPORARILY UNAVAILABLE FOR MAINTENANCE

findsite1.png
FINDSITE-metal is an extension of FINDSITE that predicts metal-binding sites, residues and binding metal preferences from evolutionarily related templates detected by threading. Furthermore, it employs machine learning to improve the prediction accuracy particularly against low-to-moderate quality protein structures. Thus, crystal structures as well as protein models can be used as the targets.


NOTE:

  • This web service is freely available to all academic users and not-for-profit institutions.
  • Commercial users wishing an evaluation copy should contact skolnick@gatech.edu .
  • Commercial users may license the FINDSITE-metal software after completing the license agreement and sending it to skolnick@gatech.edu Download the license agreement.


If you use FINDSITE-metal, please cite the following paper:

  • Brylinski M and Skolnick J (2011) FINDSITE-metal: integrating evolutionary information and machine learning for structure-based metal binding site prediction at the proteome level. Proteins: Structure, Function, and Bioinformatics. 79(3):735-751.


Submit your target structure to our server

NOTE: The server accepts only single protein chains 40-400 residues in length. Only the ATOM section is parsed, non-standard residues are removed. Missing heavy atoms are added. PROSPECTOR_3, SP3 and Sparks2 threading algorithms are used to identify evolutionarily related templates. You can upload either the crystal structure or the predicted model of your target. Estimated TM-score must be in the range of 0.0-1.0.


Target ID (2-8 characters)



Target PDB file



Estimated TM-score (use 1.0 for crystal structures)



Your e-mail address



Download the source code and install FINDSITE-metal locally


FINDSITE-metal manual (PDF)

version 1.0 latest



FINDSITE-metal sources

version 1.0 latest



PDB library

September 2010 latest
April 2010


Take a look at FINDSITE-metal benchmarks

Benchmark results provided below were obtained using weakly homologous (<35% sequence identity to the target) templates identified by three threading algorithms: SP3, Sparks2 and PROSPECTOR_3. The list of proteins and associated metal binding ions is available here. We tested FINDSITE-metal using different quality target structures:


Crystal structures

Target structures in PDB format
FINDSITE-metal output files



Structures distorted to 2Å Cα RMSD

Target structures in PDB format
FINDSITE-metal output files



Structures distorted to 4Å Cα RMSD

Target structures in PDB format
FINDSITE-metal output files



Structures distorted to 6Å Cα RMSD

Target structures in PDB format
FINDSITE-metal output files



Weakly homologous chunk-TASSER models

Target structures in PDB format
FINDSITE-metal output files


FAQ

Q: How long does it take before my results are ready?
A: The simulation time depends on several factors: the length of the protein, the number of identified templates and the workload on our computer cluster. Typically, the results should be ready within 1-2 days. We are working to reduce the turn around time.

Q: How long will my results be kept on the server?
A: The results will be kept on the server for one month.

Q: How to estimate the TM-score for my protein model?
A: This is described in FINDSITE-metal manual, see Appendix A.


This service was created and is maintained by Michal Brylinski