The Web Site
The web site enables the user to compare three-dimensional structure of a protein prediction with the real structure as discovered in the lab. We attempt to place the prediction in a way that resembles the real structure best, so those segments of high correlation will be shown close together. In order to fit these two objects in space we displace and rotate the prediction over the true structure using methods described in the project description. These provide good superposition of the two structures. For visual imaging we use the "Molscript" and "VRML" software that give three-dimensional movable visualization of the protein and its prediction.
The web site is followed by an easy to use online manual for the users' facilitation. All the user needs to do is insert two files, a prediction and a true model of the protein. To ease the use predictions may be submitted in TS or AL format. There is also direct access to the prediction files of CASP3 (critical assessment of techniques for protein structure prediction) meeting. IN case the user does not have a file containing the true protein 3D structure we automatically retrieve such from "PredictProtein" database by the proteins code.
Homology is found automatically based on our "maxsub" algorithm or by manual checking of preferred acids to a subgroup.
We supply three forms of prediction quality meters:
Numerical results that are 'cold' meters for prediction quality.
Graphical illustration of the protein and the prediction (separately), with emphasis on secondary structure.
Graphical comparison of the two by superimposing the prediction on the true protein with emphasis on amino acids distances.
On the left of the entrance page there are buttons leading to further information the site provides. The 'Help' button gives general explanations about the site, offers a manual and assistance for comprehending the output and other information relevant for the use of the site. The 'Links' button gives reference to other web sites that might interest our users. Such links include the Molscript homepage (providers of the software needed to produce protein structure visualization), the Cosmo Player homepage (providers of the VRML software needed to graphically illustrate our 3D structures), The PredictProtien homepage (where laboratory found protein structures are displayed and where users can check protein codes) and CASP3 (where many prediction files are given).
On the entrance page the user is welcomed to compare a prediction with a true protein structure. The user enters the file name of the true protein structure (using TS format) or its PredictProtien code in the specified boxes at the top of the page marked as 'protein file'. The user must also enter the file name of the prediction that is wanted to be compared to the protein (using either TS or AL format) or its code in the specified boxes marked as 'prediction file'. At the bottom part of the page the user sets the maximal distance threshold for the automated maximal subgroup selection (using maxsub) or indicates the will to pick such a subgroup manually by this group a second superposition transformation will be calculated. Pressing the submit button submits the information to the software and the result page appears.
The main result page gives immediate information about the correlation of the prediction and the protein using both kinds of comparison: the traditional based on all amino acids and ours, which is based on a selected group. The immediate information includes the names of the files compared, the RMS distance of the whole protein and of the selected subgroup and a GL score of that sub group (GL score is a meter evaluating the quality of the prediction.
The first two links are for the numerical information pages. These two provide the rotation matrix and displacement co-ordinates used to produce this transformation, the amount of amino acids selected to form the sub group, and the distances of each of the acid in the prediction file to its parallel in the actual protein. The upper link is for information using the traditional method while the lower one is information using the subgroup.
The next two links are for graphical illustrations of the protein and the prediction (given separately) introducing the three dimensional structure using VRML. Emphasis is given to secondary structure for better subjective evaluation of the prediction quality. Different domains are represented in separate colours while alpha helices and beta sheets are represented by coils and arrows respectively.
The last two links are to more graphical information showing the superposition of the prediction over the protein. These two (one for traditional method and the other for the subgroup) show the superposition after our transformation was performed and indicates distances of matching amino acids in the two structures by a colour scale. Segments of higher homology will be coloured blue while segments of lower homology (and greater amino acids' distances) will be coloured red.
Instructions for using this site
How to read the output
Filling the Form
You are welcome to put your prediction in the "Prediction File" boxes. Please specify weather the data is in AL or TS format.
You may also try a prediction from PredictionCenter by specifying it's AL code (such as T0043AL028_5).
You must also give the proteins real structure file in the "Protein File" box. The site can call such a file automaticly if you specify it's pdb remote code (such as 1hka).
Press 'submit' to start comparison.
Comprehending the Output
The output page brings 6 possabilaties:
* 'Results for transform using all acids'
* 'Results for transform using 'maxsub' subgroup'
* 'View actual protein image (VRML)'
* 'View predicted protein image (VRML)'
* 'View superposition of both predicted and actual protein'
* 'View superposition of both predicted and actual protein using 'maxsub''
The first 2 give numerical results, inicluding RMS distance, GL score, transformation rotation matrix, transformation displacement, and offers a link to a list including each of the amino acids of the prediction, and its distance from the real one, after the transform.
The second 2 produce 3-dimensonal graphic illustration of proteins' secondery structure. Alpha helicies and beta-sheets are conviniantly marked by coils and arrows respectively. Differant domains are colored in seperate colors.
The last 2 produce 3D illustration of the superpositioning of the prediction over the actual protein. Distances are visualized by colors, the bluer the better. The first superposition is after using the traditional method (concidering all acids), while the second one is transformed using only a subset of the acids, either picked manualy or by 'maxsub'.