The IntFOLD server is a fully integrated pipeline, combining each of our cutting edge tools for the prediction of structure and function from a single sequence and is intended for use by both expert and non-expert biologists alike. A user-friendly interface is provided for query sequence submission, which allows non-expert users to predict a variety of protein structural features.
The IntFOLD server provides a unified interface for:
- Tertiary structure prediction/3D modeling with built-in estimates of model accuracy (EMA)
- 3D model quality estimates with the option to refine/fix errors
- Intrinsic disorder prediction
- Domain prediction
- Prediction of protein-ligand binding residues
- Protein structural domain boundaries
- Natively unstructured or disordered regions in proteins
- Determination of protein-ligand interactions
The component methods have been independently evaluated via the successive blind CASP experiments and the continual CAMEO benchmarking project. The IntFOLD server has established its ranking as one of the best performing publicly available servers, based on independent official evaluation metrics.
A single amino acid sequence for the protein chain is the only required input for the server. A query protein sequence is first submitted, with ∼40 new models generated. Secondly, the models are fed into the ModFOLDclust2 model quality assessment algorithm, which ranks the models by model quality. The models are then used along with a combination of local error and template information to produce all the resulting output for 3D structure prediction (TS), domain prediction (DP), binding site residue prediction function prediction (FN), disorder prediction (DR) and model quality assessment (QA).
However, users also have the option to provide a short memorable name for their prediction job and an email address, which will only be used to provide a notification of the link to the results when the predictions are completed. If users do not wish to be notified via email, then they can bookmark the link to the results page for later viewing.
The graphical output is presented as a single table that graphically summarizes all prediction data using thumbnail images of ASE plots and models, links to the template information and colour coded scoring. It is always recommended to choose the model with the highest score or lowest P-value. The confidence rating relates to the P-value. For example, a ‘CERT’ rating relates to models where P < 0.001, i.e., less than a 1/1000 chance that the model is incorrect. So all ‘CERT’ models are highly likely to have the correct fold. However, the models with the lowest P-values are more likely to have the highest backbone accuracy and overall quality. Several new user interface options are available. Users have the option to download coordinates and view the detailed IntFOLD5-TS tertiary structure prediction results interactively in 3D and submit individual 3D models for further refinement using ReFOLD via simple push buttons. Downloadable coordinates and interactive 3D views of the protein ligand interactions can also be accessed via the FunFOLD results summary image. In addition, clicking on the DISOclust disorder prediction profile images and the thumbnail images of the ASE score profiles from ModFOLD7_rank will allow users to view and/or download higher quality versions of the plots.