In order to submit the job you need to provide:
Mark the option “Do not show my job on the queue page” if you don’t want the job to be visible to anyone else on the queue page (QUEUE). The example of submitting the job named “example_1SUR” by providing the PDB code: “1SUR” is displayed below.
After clicking the Submit button, the following info will appear, which contain the unique link to your job:
If you have chosen “Do not show my job on the queue page” it is important to save the link to your job, otherwise your job will be accessible from the queue page: ( QUEUE).
Under the unique link to your job you’ll find the job status updates and finally the job results.
The job status information will start from the “in queue status”:
Under the job link, the results are accessible from the Models menu tab (marked below in a red circle). Under the Models tab, results are divided into several sub-tabs (marked below in green circles):
To see description of the displayed pictures, or the data to download (accessible under the blue buttons), drag the cursor over a picture, or a button.
The Multimodel subtab presents the following data:
Residue fluctuation profile shows relative propensities of protein residues to deviate from an average dynamics (trajectory) structure. The CABS-flex employs simulation procedure from the work of [ ref 1], where it was demonstrated, that the average Spearman’s correlation coefficient for the fluctuations along protein chains between CABS and the atomic MD (all-atom, explicit water, for all protein metafolds using the four most popular force-fields) is average on the level of 0.7 (which has been confirmed in further test studies). Importantly, this level of correspondence is similar to that of between different MD force-fields.
The residue fluctuation values are also included into the final PDB output models (the temperature factor column, 61 – 66 columns in the PDB file, is replaced with the fluctuation values that can be visualized as colors using standard molecular visualization software).
The superimposition is done by the Theseus application.
The picture is shown to provide at a glance visual information about the structural ensemble of predicted models. Note that, the models are displayed in a random orientation. In order to create your own customized molecular visualization picture, click the button below: "Multimodel [PDB]" and visualize the downloaded file with a molecular visualization software of your choice. See also Gallery of protein structure fluctuations resulted from the CABS-flex server.
The Model 1, Model 2 and each subsequent Model subtabs presents the following data:
The picture is shown to provide at a glance visual information about the similarity of the predicted model to the input structure. Note that, the structures are displayed in a random orientation. In order to create your own customized molecular visualization picture, click the button below: "Download model with superimposed input structure"). See also Gallery of protein structure fluctuations resulted from the CABS-flex server
Under the Details tab, results are divided into several sub-tabs:
Clustering of protein models is the task of separating a set of protein models (here a protein dynamics trajectory) into groups (called clusters). The clustering is done in such a way that models are more similar in the same group to each other (here in the sense of RMSD measure), than those in other groups (clusters). CABS-flex utilizes classical K-means clustering method.
After clustering is done, each cluster representative is chosen (always the model which average dissimilarity to all models in a cluster is minimal). Predicted protein models, presented in the Models tab, are each cluster representatives (the clusters and the corresponding models are marked by the same numbers, e.g. Model 1 represents Cluster 1).
The clusters are numbered/ranked according to cluster density values, from the most dense (numbered as a first) to the least dense one.
The subtab Clustering Data contains a table with the following clusters data:
In the example above, the Cluster 1 contains 357 models (selected out of entire trajectory which contains 2000 models), whose average RMSD between all pairs of models in the Cluster is 1.2 Angstroms, and the Cluster density is 300 (note that the average cluster RMSD value given in the table is rounded to one decimal place, however in the calculation of the cluster density value the exact number is used).
Since the Cluster 1 is the most dense and most numerous, the Model 1 can be considered as the representative of the most dominant conformation in the entire fluctuation ensemble, followed by the Model 2 (representative of the second most dominant structure), and so on.
The table contains RMSD and GDT_TS values (calculated on the Cα atoms) between the predicted models and the input structure. Note that GDT_TS metric is intended as a more accurate measurement than the more common RMSD.
Read more about the root-mean-square deviation (RMSD) measure
Read more about the global distance test (GDT, also written as GDT_TS to represent "total score") measure.
The table contains RMSD values (calculated on the Cα atoms) between the predicted models.
Read more about the root-mean-square deviation (RMSD) measure.
The table contains GDT_TS values (calculated on the Cα atoms) between the predicted models.
Read more about the global distance test (GDT, also written as GDT_TS to represent "total score") measure.
The Theseus simultaneously superimposes multiple protein structures and finds the optimal solution to the superposition problem using the method of maximum likelihood. By downweighting variable regions of the superposition and by correcting for correlations among atoms, the maximum likelihood superpositioning method produces much more accurate results than conventional methods using least-squares criteria. Read more [ ref 2], Theseus website.
CABS-flex requires input PDB files with continuous (without breaks) protein chain. PDB files with gaps in structure have to be first prepared by filling up the missing fragment. Below is the list of example software and on-line servers that enable filling in the gaps in incomplete 3D models:
© Laboratory of Theory of Biopolymers, Faculty of Chemistry, University of Warsaw 2013