POLYVIEW-3D Tutorial

Part 5. Protein structure annotation settings

These options include some of the most complex types of queries and structural analysis provided by POLYVIEW-3D. For example, they enable performing various structural and functional analyses, using both in house software, such as SPPIDER for the recognition of protein interaction interfaces, as well as several external servers, including CASTp for structural pockets identification, and ConSurf for assessment of evolutionary conservation. The results of these external resources are processed further (to a different degree) and combined with additional analysis performed by POLYVIEW-3D. One example is the ability to combine identification, mapping and prediction of protein interaction interfaces performed by SPPIDER with the analysis of putative pockets (and other topographical features) identified by CASTp. See sections below for details.

There are two general groups of settings within this options set. The first group invokes automatic requests to other web-servers and databases to obtain additional structural annotations. The second group uses precomputed results that were previously received (and possibly processed) from other resources. In both cases, the resulting data is further processed by the POLYVIEW-3D server in order to yield 3D images of the macromolecular structures overlaid with the requested annotations, as described below.

Finding pockets using CASTp

Invoking this option results in a request sent automatically to the CASTp server in order to perform a search for structural pockets within a given query protein. Such identified pockets can be filtered by their area or volume. These putative pockets are also automatically colored according to the CASTp color convention and characterized in tabular form in terms of both atoms and residues they comprise. Pockets can also be optionally overlaid with the prediction of interacting sites automatically performed using the SPPIDER server. Pockets overlapping with predicted interfaces may represent interesting targets for drug design and docking simulations.

Below is an example of a request to find pockets using CASTp, with pocket area cutoff of 100Ų. The structure of purine nucleoside phosphorylase in transition state was taken as a query structure (PDB id 1b8o).

Pockets larger or equal 100Å2
Click on respective image to see options used for its rendering.

Generated image is accompanied by annotation table with the lists of residues located at each of the pockets.

Determination of domains using Pfam

When this option is selected, POLYVIEW-3D performs a sequence homology search using BLAST against local version of Pfam database. Amino acid sequence used for query is derived from ATOM section of the PDB file. If significant sequence homology (E-value ≤ 0.001 AND sequence identity ≥ 70%) is found to one or more domains, they are mapped to the queried structure using distinct colors for each domain. The same structure as in previous example was used as a query (PDB id 1b8o) in the image below.

Domains from Pfam
Click on respective image to see options used for its rendering.

Generated image is accompanied by annotation table with the lists of residues found to belong to different domains.

Determination of putative TM segments using PDBTM

To process the request with this option chosen, POLYVIEW-3D makes search in local copy of the PDBTM database of putative trans-membrane segments. Depending on the source of the protein structure the server either seeks the data by the PDB code or performs a sequence homology search using BLAST. In the latter case, amino acid sequence used for query is derived from ATOM section of the PDB file. If significant sequence homology (E-value ≤ 0.001 AND sequence identity ≥ 70%) is found to one or more proteins, the best matching homolog is taken. Images below demonstrate the membrane spanning segments annotation as determined in PDBTM using the TMDET automated algorithm (for details, please refer to the original papers). The structure of Catalytic Core (Subunits I and II) of Cytochrome c oxidase from Rhodobacter sphaeroides serves as an example of alpha-helical TM protein (PDB id 2gsm), whereas the structure of the sucrose-specific porin ScrY (PDB id 1a0s) from Salmonella typhimurium represents beta-barrel TM proteins.

TM segments from PDBTM Alpha helical	TM segments from PDBTM Beta barrel
Click on respective image to see options used for its rendering.

Generated image is accompanied by annotation table with the lists of residues computationally determined to be membrane spanning segments.

Mapping interacting sites from PDB complexes using SCORPPION

With this option, one can retrieve information about all interacting sites found among close sequence homologs deposited in PDB. Search is automatically performed using the SCORPPION server. Interacting sites are the residues in contact with other protein chains, DNA or RNA, or ligands. Different types of interactions, as well as their combinations are colored distinctly. Interacting sites can be optionally contrasted with protein interfaces predicted using SPPIDER.

Below is an example of mapping different types of interactions from close sequence homologs to a query structure of the transcription factor CSL (PDB id 1ttu, chain A).

Interfaces by SCORPPION
Click on respective image to see options used for its rendering.

Generated image is accompanied by annotation table with the lists of residues found to have different types of contacts.

Predicting protein interface using SPPIDER

This setting belongs to the second group of options that utilize pre-computed results, which were received prior to submitting a query to POLYVIEW-3D. The SPPIDER server optionally generates a PDB file with temperature factor fields modified according to the probability of that residue being involved in protein-protein interactions. These files can be submitted to the POLYVIEW-3D server directly, as custom PDB files with chains of interest to be colored using By B-factors color scheme (see Chains rendering settings).

Examples below illustrate the use of POLYVIEW-3D in this case, with interacting sites found in the CSL transcription factor (PDB id 1ttu, chain A) using SPPIDER. The predictions are encoded using both classification and regression approach (with a binary class assignment in the first case, and a probability of being within an interaction interface in the latter case, respectively), and incorporated in the corresponding B-factor fields. Color scheme applied is By B-factor and described in Chains rendering settings section. PDB files modified by the SPPIDER server and used for the images are also available for downloading (classification-based and regression-based prediction).

Interfaces by SPPIDER as classification	Interfaces by SPPIDER as regression
Click on respective image to see options used for its rendering.

Generated images go along with annotations where residues are grouped by either classes or probability bins (see corresponding tables below).

Finding functional regions using ConSurf

The ConSurf server ranks residues according to their relative conservation scores that can be used for the identification of putative functional regions within a given structure. The server also modifies the original PDB file to encode conservation scores by replacing B-factor fields. This kind of file can be directly submitted to POLYVIEW-3D. By specifying chain of interest to be colored by B-factors, one can obtain high-quality images of ConSurf results with its original coloring scheme.

Below is an example of ConSurf output, with conservation scores encoded as B-factors and residues colored using the original color scheme. The same protein was used as in the previous section (PDB id 1ttu, chain A). Color scheme applied is By B-factor and described in Chains rendering settings section. PDB file modified by the ConSurf server and used for the image is also available for download.

Functional regions by ConSurf
Click on the image to see options used for its rendering.

Generated image is accompanied by annotation table with the lists of residues found to have different types of contacts.

Analysis of docking models from CAPRI

POLYVIEW-3D provides an option for analyzing the docking models in the CAPRI format obtained by different programs or web-servers for protein-protein docking. In addition to visualization of docking models, POLYVIEW-3D performs their analysis and assessment. In particular, when a custom PDB file with multiple models of a protein complex (e.g., generated by the ClusPro server) submitted to POLYVIEW-3D using the Docking models option, it triggers SPPIDER predictions to perform for all chains in the model. Unbound structures of these chains are used to predict putative interaction interfaces, which are then compared to interfaces observed in each model. The fraction of residues within the interface in a given model that overlaps with SPPIDER predictions (averaged over both chains) provides a simple score to rank these models. In addition, the surface area and average hydrophobicity for each interface within these models are computed to provide a basis for further analysis and visualization. Models can be then re-ranked according to these values and sent to refine image rendering.

The table below includes the output of POLYVIEW-3D assessment of docking models for the system used as CAPRI target #9, and described in more details in Example 2 of the Advanced examples section. The two chains were submitted to the ClusPro server, in order to obtain 10 best ranking models of the protein complex (download results). The option for models assessment has been invoked by requesting the overlay with protein interfaces predicted by SPPIDER. Note that table rows (and thus different models) can be easily ordered (re-sorted) according to measures included in the columns. In addition, a form is provided to alter the sensitivity used to identify interaction sites from the SPPIDER initial (default) definition. It allows the user to make the interface overlap measure either more restrictive or loose.

Analysis of docking models (Click on the header of column to re-sort models)

Model	Image	(1) ASA, Å2	(2) HP index	(3) Overlap, %	Chain A Interface ASA, Å2	Chain A Interface HP	Chain B Interface ASA, Å2	Chain B Interface HP	Residues at the interface (Overlap with SPPIDER prediction, %)
1		2436	1.04	52.42	1246	1.06±0.72	1190	1.01±0.72	Chain: A (54.84%) Overlapped: 2 3 5 6 7 9 ... Observed: 21 37 46 63 112 ... Predicted: 1 4 8 12 13 14 ... Chain: B (50.00%) Overlapped: 2 3 5 6 7 9 ... Observed: 21 37 38 46 ... Predicted: 1 4 8 12 13 14 ...
2		2150	0.72	45.69	1081	0.68±0.67	1069	0.75±0.67	Chain: A (41.38%) Overlapped: 1 2 3 6 41 ... Observed: 34 36 37 38 112 ... Predicted: 4 5 7 8 9 10 ... Chain: B (50.00%) Overlapped: 1 2 3 6 41 ... Observed: 34 36 37 38 ... Predicted: 4 5 7 8 9 10 ...
3		2438	0.95	32.73	1218	0.95±0.73	1220	0.95±0.74	Chain: A (32.14%) Overlapped: 5 8 9 12 49 53 59 61 134 Observed: 62 63 65 66 ... Predicted: 1 2 3 4 6 7 ... Chain: B (33.33%) Overlapped: 5 8 9 12 49 53 59 61 134 Observed: 62 63 65 66 67 96 97 ... Predicted: 1 2 3 4 6 7 ...
4		2091	0.96	47.22	1050	0.98±0.67	1041	0.95±0.67	Chain: A (44.44%) Overlapped: 3 7 10 11 15 17 18 40 Observed: 21 25 28 29 32 ... Predicted: 1 2 4 5 6 8 9 ... Chain: B (50.00%) Overlapped: 3 7 10 11 15 17 ... Observed: 21 25 28 29 32 ... Predicted: 1 2 4 5 6 8 9 ...
5		2374	1.02	21.45	1176	1.00±0.74	1198	1.03±0.73	Chain: A (22.22%) Overlapped: 53 59 61 134 137 139 Observed: 62 63 65 66 67 69 ... Predicted: 1 2 3 4 5 6 7 8 ... Chain: B (20.69%) Overlapped: 53 59 61 134 137 139 Observed: 62 63 65 66 67 69 ... Predicted: 1 2 3 4 5 6 7 8 ...
6		2479	0.83	48.34	1253	0.88±0.69	1226	0.79±0.69	Chain: A (51.52%) Overlapped: 1 2 3 6 7 40 41 ... Observed: 21 28 31 32 33 34 ... Predicted: 4 5 8 9 10 11 12 ... Chain: B (45.16%) Overlapped: 1 2 3 6 10 40 146 ... Observed: 31 32 33 34 36 37 ... Predicted: 4 5 7 8 9 11 12 ...
7		1276	0.79	46.05	634	0.80±0.69	642	0.77±0.70	Chain: A (50.00%) Overlapped: 1 2 3 5 6 10 41 ... Observed: 36 161 163 174 177 ... Predicted: 4 7 8 9 11 12 ... Chain: B (42.11%) Overlapped: 1 2 3 5 6 10 ... Observed: 36 161 162 163 ... Predicted: 4 7 8 9 11 12 ...
8		2305	1.02	30.81	1165	1.00±0.69	1140	1.04±0.69	Chain: A (29.63%) Overlapped: 6 53 56 58 59 61 134 139 Observed: 62 63 64 66 67 69 ... Predicted: 1 2 3 4 5 7 8 9 ... Chain: B (32.00%) Overlapped: 6 53 56 58 59 61 134 139 Observed: 63 64 66 67 114 115 ... Predicted: 1 2 3 4 5 7 8 9 ...
9		2081	0.89	51.47	1056	0.86±0.71	1025	0.91±0.70	Chain: A (50.00%) Overlapped: 144 147 151 154 165 ... Observed: 150 158 162 163 164 ... Predicted: 1 2 3 4 5 6 7 8 9 ... Chain: B (52.94%) Overlapped: 144 147 151 154 165 ... Observed: 150 158 162 163 164 ... Predicted: 1 2 3 4 5 6 7 8 9 ...
10		2059	1.12	51.19	1032	1.11±0.67	1027	1.13±0.66	Chain: A (52.38%) Overlapped: 3 6 7 10 11 15 17 ... Observed: 21 25 28 29 32 33 ... Predicted: 1 2 4 5 8 9 12 13 ... Chain: B (50.00%) Overlapped: 3 6 7 10 11 15 17 ... Observed: 21 25 28 29 31 32 33 ... Predicted: 1 2 4 5 8 9 12 13 ...

¹ − Total solvent accessibility buried upon complex formation
² − Mean hydrophobicity index of the interfaces
³ − Average overlap across chains between observed interfaces in a given model and predicted by SPPIDER
Annotation of residues:
Overlapped(Red) - observed and predicted to be at the interface
Observed(Blue) - located at the interface in a given model, but not predicted
Predicted(Yellow) - predicted to be at the interface, but not observed in a model

Animation of macromolecular movements

There are a number of programs and web-servers performing simulation and analysis of macromolecular movements, for example Analysis of Dynamics of Elastic Network Model (AD-ENM) or Database of Macromolecular Motions (MolMovDB). Motion trajectories and structural distortions may help identify flexible and rigid regions, localize hinges and stable domains. In this regard, POLYVIEW-3D has a function that analyzes trajectory quantitatively and produces both animations and 2D-trajectory images. In order to invoke this function, the user has to chose the option Trajectories and distortions and submit a PDB-formatted coordinate file with conformational changes recorded as models.

When this structure annotation is requested, POLYVIEW-3D ignores a current setting of the Type of request option in the Image Settings field set. In contrast to the option that animates Models from the Animation settings, this option generates a movie comprising of all available models and makes the motions reversible. Images shown below represent two macromolecular motions as deduced at the corresponding servers. Coordinate files were taken from the examples available at those servers and can be downloaded here: AD-ENM example of myosin deformation, MolMovDB example of calmodulin hinge motion.

Low-frequency motion by AD-ENM	Morph by the Morph Server
Click on respective image to see options used for its rendering.

Along with animation, the resulting page presents 2D-plots with trajectories of structural changes in terms of secondary structure (SS) and relative solvent accessibility (RSA), and with per-residue analysis of protein flexibility. The details on the measures used to quantify conformational changes as well as the options to customize a 2D-plot can be found in the POLYVIEW documentation. Below are 2D-plots produced for SS and RSA changes occurring during the calmodulin hinge motion as interpolated by the Morph Server.

Conformational changes in calmodulin in terms of SS
Conformational changes in calmodulin in terms of RSA

Last update: October 2009

<< Prev Next >>