The ModFOLD Model Quality Assessment Server

About the server

The ModFOLD server can provide:

1. A single score and a p-value relating to the predicted quality of a single 3D model of a protein structure.
2. Rankings for multiple 3D models for the same protein target according to predicted model quality.
3. Predictions of the local quality (per-residue error) within multiple models.

Figure 1 taken from McGuffin (2008). The ModFOLDclust predicted per-residue error (left) for an example model is compared to the observed error obtained from the alignment to the native structure (right). Each image was rendered using Pymol (http://www.pymol.org). The colours represent the residue accuracy according to the temperature scheme (blue indicates residues closest to the native structure; red, those furthest from the native structure).

ModFOLD version 2.0

ModFOLD version 1.1

News

• Nov 2008: ModFOLD version 2.0 (beta) online for testing - the method aims to combine the ModFOLD and ModFOLDclust methods and outputs results in QMODE2 format.
• July 2008: Fair usage policy introduced due to high demand.
• March 2008: The ModFOLD servlet code has been moved to a new server for CASP8.
• Feb 2008: If models are incorrectly numbered then the server will now attempt to automatically renumber the ATOM records in each model in order to match the residue positions in the sequence.
• Jan 2008: The ModFOLD server paper has been published in Bioinformatics.
• Dec 2007: The DISOclust disorder prediction method has been intergrated into server. Click here for further info.
• Oct 2007: Version 1.1 of the ModFOLD server is now online. Two programs are now available:
1. ModFOLD v1.1 - a fast true Model Quality Assessment Program (MQAP) that works for single or multiple models.
2. ModFOLDclust v1.1 - a slower clustering method that works for multiple models only, but also provides per-residue local quality assessement.

References

If you have used the ModFOLD server to generate predictions, then please cite to the following paper:
• McGuffin, L. J. (2008) The ModFOLD Server for the Quality Assessment of Protein Structural Models. Bioinformatics, 24, 586-7. PubMed
If you want to refer to the latest description of the ModFOLDclust method and ModFOLD v 2.0, then please cite the following paper:
• McGuffin, L.J. (2009) Prediction of global and local model quality in CASP8 using the ModFOLD server. Proteins: Structure Function and Bioinformatics, In press.
If you want to refer to the original description of the ModFOLD method, then please cite the following paper:
• McGuffin, L. J. (2007) Benchmarking consensus model quality assessment for protein fold recognition. BMC Bioinformatics, 8, 345. PubMed

Please also refer to the following papers which describe MODCHECK and ProQ:

• Pettitt, C. S., McGuffin, L. J. & Jones, D. T. (2005) Improving sequenced based fold recognition by use of 3D model quality assessment. Bioinformatics, 21, 3509-3515. PubMed
• Wallner, B. & Elofsson, A. (2003) Can correct protein models be identified? Protein Sci. 12, 1073-1086.

Versions

Overview ModFOLD v 1.1 server

The server now includes two different methods: ModFOLD and ModFOLDclust.
The new version of ModFOLD now combines a total of 6 scores using a neural network (the 4 scores from the previous version plus two addtional secondary structure measures).
The ModFOLDclust method includes clustering of multiple models providing higher accuracy and residue error scores for the local quality.

Overview of the original ModFOLD method

The original method combined the results from ModSSEA - a model quality assessment program (MQAP) based on secondary structure element alignments - with several established methods. Data from ModSSEA, MODCHECK, ProQ-MX and ProQ-LG are combined using a neural network which has been trained using TMscores.