Fit3D Webserver

Terms of Service

Here you can find detailed explanations on how you can use our web service to search for spatial residue patterns in proteins. You can easily click on the respective category to show the instructions. Brief answers to most usage-related questions are given in the FAQ section .

How does it work?

The search algorithm behind Fit3D is basically a combinatorial search approach that works by evaluating two constraints: residue composition and geometric similarity (root-mean-square deviation/RMSD) of match candidates. In the following the basic algorithmic procedure is elucidated:

determination of the maximal spatial extent of the query motif
determination of the residue composition of the query motif
for all search target structures, do
- eliminate residues not element of the query motif residue composition
- combine sets with the size of the query motif, similar spatial extent and matching residue composition (match candidates)
- determine RMSD for match candidates

Pairwise distance filtering can be used to eliminate match candidates with inter-residue distances not compatible to the query motif. For detailed information on how the algorithm works and why it is a valuable structural motif search method, please refer to Kaiser et al. 2015, .

Prerequirements

Before you can start the large-scale screening for a structural motif you will need:

the query structural motif in PDB format (you can also the our extraction wizard or our API to extract a query motif)
(optional) a list of search targets as text file (you can also use our predefined lists)

download archive of predefined lists used by Fit3D

Extracting a motif

To define a new structural motif you can use our extraction wizard that allows convenient and intuitive motif definition. You can either access a PDB structure by its PDB-ID or you can upload a structure in PDB format. The video below shows the extraction process step-by-step using a known PDB structure:

Choose "Extract motif" in the navigation menu on the left side.
Enter a valid PDB-ID or upload your own structure.
On the next wizard page: select residues that should be part of the query motif using the drop-down menu.
After clicking the button "Extract", the motif structure is visualized and the toolbox shows further details.
Directly submit a new search job by clicking "Submit" or download the motif structure for offline usage (click "Download")

Note: If the extracted motif is rated as too complex to be processed by the web server, the command line version of Fit3D can be used.

Submitting a job

To submit a new job click "Submit job" in the left navigation menu. A motif file in PDB format is mandatory and must be provided by the user. Otherwise, the extraction wizard can be used to obtain a query structure. The target search space can be customized by uploading a list of PDB-IDs separated by line break. Alternatively, different pre-defined target lists, that were computed based on sequence or structure homology analysis, can be selected. The video below shows the submission process in detail:

Choose "Submit job" in the navigation menu on the left side.
Upload a query motif structure in PDB format by clicking "+motif file".
Select a predefined target list of search targets or upload your own target list by clicking "+target list".
Choose a RMSD limit up to which matches will be reported.
(optional) Enter a representative description and an email address in order to be notified after your job has finished.
(optional) Use the "Advanced settings" tab to enable/disable pairwise distance filtering, to change the p-value calculation method, to select alignment atoms, or to define position-specific exchangeDefinitions (PSEs).

Advanced settings

Position-specific exchangeDefinitions (PSEs) for the motif search can be easily defined by using the advanced settings dialog. This can be useful for analysis of diverse protein families or mutagenesis experiments. PSEs are residue substitutions that are allowed during motif matching. As the name suggest, Fit3D handles position-specific residue substitutions, e.g. one can define a single residue in the query motif to be also matched against other residue types. The video below shows how to allow matching of different amino acid types:

Activate PSEs for your query motif by clicking the "Define exchangeDefinitions" button (important: the query motif must be already known). Select possible substitutions for single residues from the drop-down menu. Click "Set exchangeDefinitions" to apply your selection.
Note: The total number of position-specific exchangeDefinitions (PSEs) is restricted to three.

View results

To view the results of your job click "View results" in the left navigation menu. You can now select a corresponding job to show all detailed results. Watch the video below to see the whole process:

All jobs submitted during the active browser session are accessible under the "View results" section.
If you used email notification you can recover all results of the session later. Important: the results of each job are deleted from our server after 72 hours.

Result page layout

When viewing the results of your job, motif matches are ranked based on the lowest least-root-mean-square deviation (RMSD). Hints on how to understand and interpret the RMSD are given in the FAQ section. If you want to investigate a single match in detail, just click on "Show" or "Show in structure". Furthermore, you can directly access the corresponding Protein Data Bank (PDB) entry of the structure. Note that you can download all results or the RMSD distribution plot by clicking on the "Action" drop down menu on the toolbar at the very top of the result page.

p-value theory

Beside geometric similarity matches are rated according their statistical significance. Every reported match to a query motif is labeled with a p-value such that the user can assess true positives intuitively. However, a significant p-value does not necessarily indicate a true positive match because the underlying models are based solely on the distribution of RMSD values. Hints on how to understand and interpret the p-value are given in the FAQ section. No biological features are represented by those models. To estimate statistical significance, Fit3D implements two statistical models:

kernel density based model, corrected with a maximum likelihood estimated point-weight (according to Fofanov et al. 2008 )
empirically derived model according to Stark et al. 2003

Illustration of a statistical model to estimate significance of local similartities in protein structures according to Fofanov et al. 2008. A kernel density esimation of the match RMSD distribution is calculated using the Sheather-Jones bandwidth selection algorithm. Additionally, a point-weight correction (pwt) is applied to estimate matches beyond the RMSD limit by a maximum likelihood approach.

Preprocessing

Motif search

Others