SCOP2 Documentation

1. Overview.

1.1. Database description.

1.2. Definitions.

1.2.1. SCOP2 categories.

1.2.1.1. Protein types.

1.2.1.2. Evolutionary events.

1.2.1.3. Structural classes.

1.2.1.4. Protein relationships.

1.2.1.4.1. Evolutionary relationships.

1.2.1.4.2. Structural relationships.

1.2.1.4.3. 'Other' relationships.

1.2.2. SCOP2 domains.

1.3. SCOP2 representative sequences and structures.

1.4. SCOP2 controlled vocabulary ( keywords and tags ).

2. Getting started. Browsing and searching SCOP2.

2.1. SCOP2-graph.

2.1.1. SCOP2-graph Main window.

2.1.2. SCOP2-graph Node popup window.

2.1.2.1. Structure tab.

2.1.2.2. Parents and Children tabs.

2.1.2.3. Comment tab.

2.1.3. SCOP2-graph Configuration window.

2.1.4. How to search SCOP2-graph.

2.2. SCOP2-browser.

2.1.1. Representative domain page.

2.1.2. Cross-Reference Map.

2.1.3. Clustered Domains.

2.1.4. How to search SCOP2-Browser.

3. Molecular viewers.

3.1. Jmol.

3.1.1. Basic commands.

3.2. Rasmol and rmscop configuration.

3.2.1. Installing Rmscop.

4. Linking to SCOP2.

5. SCOP2 identifiers.

6. Download.

7. SCOP2 REST web-service.

7.1. Making a REST request.

7.2. Example request.

7.3. Content Type.

8. Acknowledgements.

 

1. Overview.

SCOP2 is a successor to the Structural Classification of Proteins (SCOP) database. Similarly to SCOP, the main focus of SCOP2 is to organize structurally characterized proteins according to their structural and evolutionary relationships.

1.1. Database description.

The main focus of SCOP2 is on knowledge-based expert analysis and classification of proteins that are structurally characterized and deposited in the Protein Data Bank, PDB. Proteins are organized according to their structural and evolutionary relationships, but, in contrast to SCOP, instead of a simple tree-like hierarchy these relationships form a complex network of nodes. The classification of proteins is described in terms of a directed acyclic graph in which each node defines a relationship of particular type and is exemplified by a region of protein structure and sequence. Importantly, there can be more than one parental node for a child node that allows multiple routes to a particular relationship.

1.2. Definitions.

The relationships in SCOP2 fall into four major categories: Protein types, Evolutionary events, Structural classes and Protein relationships. The first two categories do not have counterparts in SCOP.  

1.2.1. SCOP2 categories.

1.2.1.1. Protein types.

In the Protein types category, proteins are divided into four main types: soluble, membrane, fibrous and intrinsically disordered; each type to a large extent correlates with characteristic sequence and structural features. Classification of proteins according to their types resolves the inconsistent classification of membrane and coiled coil proteins, previously organised in their own classes, and allows placing them in the correct structural classes. Proteins belonging to different types, e.g., soluble and membrane proteins, can now share a common fold and even can be homologous to each other. For example, the membrane OMPA-like superfamily belongs to the same (8,10)-beta barrel fold as the Streptavidin-like proteins.

1.2.1.2. Evolutionary events.

The Evolutionary events category is introduced to facilitate the annotation of various structural rearrangements and peculiarities that have been observed amongst related proteins and which have given rise to substantial structural differences. Unlike mutations and insertion or deletions of individual residues, the events are more dramatic perturbations of protein sequence and structure. The structural features resulting from these events are inherited by the descendants of the affected protein. This category also excludes simple common events, like gene fusions, unless they are accompanied by the reorganization of domain structures by segment swapping, deletions of redundant parts, etc. An example is Duplication-fusion-deletion in the Hydroxyisobutyrate dehydrogenase/6-phosphogluconate dehydrogenase-like family.

1.2.1.3. Structural classes.

Third of the four major categories in SCOP2, the Structural classes, organizes protein folds according to their secondary structural content. They now strictly correspond to the structural classes previously defined by Levitt and Chothia. These include all alpha and all beta proteins, containing predominantly, respectively, alpha-helices and beta-strands, and ‘mixed’ alpha and beta classes, (a/b) and (a+b), with, respectively, alternating and segregated alpha-helices and beta-strands. In addition, the class of ‘Small proteins’ with little or no secondary structures is inherited from old SCOP. In SCOP2, the structural class becomes an attribute of protein fold and therefore is independent on the protein evolutionary relationships. Similar protein folds but with different secondary structural content are placed into different classes. Example: homeodomain-like and “winged” helix DNA-binding domains.

1.2.1.4. Protein relationships.

The fourth main category in SCOP2, the protein relationships, consists of three subcategories: Structural, Evolutionary and ‘Other’ relationships.

1.2.1.4.1. Evolutionary relationships.

SCOP2 retains the evolutionary levels of SCOP, Species, Protein, Family, and Superfamily but their content and definitions are different.

Species corresponds to the individual gene product and is represented by its full-length sequence, for example: Tm_MurC. A notable exception is viral polyproteins, for which protein species correspond to mature posttranslational products.

Protein groups together orthologous proteins and is defined as a subsequence that can be found on its own. In general, at this level the boundaries of the relationship are the same as in Species. The exceptions are fusion proteins found in some organisms, the Protein parts of which correspond to stand-alone proteins in other biological species. An example is Bifunctional protein Species GAL10 from yeast, Sc_GALM that is split into two proteins: Galactose mutarotase, GALM and Uridine diphosphogalactose-4-epimerase (UDP-galactose 4-epimerase), GalE.

Family corresponds to the conserved sequence region shared by closely related proteins. The boundaries of a family domain region are frequently similar to those derived by sequence family databases, e.g., Pfam, but they are adjusted to include the first (N-terminal) and the last (C-terminal) secondary structure elements in the common structural core.

Superfamily is represented by the common structural region shared by different protein families. The Superfamily domain spans from the beginning of the first common secondary structural element to the end of the last common conserved structural element. Importantly, the domains representing Family and Superfamily levels can span over more than one structural domain. Moreover, the Family region may contain more than one non-overlapping Superfamily regions.

In addition to the above levels, SCOP2 contains a new level, Hyperfamily. A Hyperfamily domain is represented by a common region shared by different superfamilies. Unlike a Superfamily region, it can be smaller than a structural domain, as it does not include the superfamily-specific regions. The Hyperfamily level is introduced mainly to deal with the most populated and structurally diverse superfamilies.

One of the principal differences between SCOP and SCOP2 is that there is no obligatory annotation for each protein at each of these distinct evolutionary levels. For example, for a protein domain that has no close ‘sequence’ relatives, but has a more distant ‘structural’ relative, there will be a Superfamily domain but no Family domains ; Winged helix DNA-binding domain of Pur operon repressor (PurR).

1.2.1.4.2. Structural relationships.

Due to the constraints of the old SCOP scheme in which structural relationships reside at the top of hierarchy, homologous proteins were classified into the same fold even if they have different folds. In SCOP2, the protein structural and evolutionary relationships are presented in separate branches (categories) to ensure more consistent classification of evolutionary related but structurally distinct proteins.

Fold in SCOP2 is an attribute of structural domain and is defined strictly on the basis of global structural features as originally described. These features are: the composition of secondary structures in the domain core, their architecture and topology. If the structural domains of evolutionary related proteins differ by any of these parameters, they can be classified into different folds.  Example: the SDR-type extended Rossmann fold differs from the Canonical Rossmann fold by an extra beta-alpha unit at the C-terminus.

IUPR (intrinsically unstructured protein region) is a new category of structural relationships, not available in SCOP. It annotates the protein regions that do not adopt globular folded structure.They are present in some of the experimentally determined structures where they adopt an ordered conformation upon binding to other macromolecules. IUPRs may occur in both individual proteins and protein sequence families. Example: the RBP-Jkappa-associated module (RAM) family is classified into the Extended binder IUPR.

1.2.1.4.3. 'Other' relationships.

The 'other' relationships, a category unique to SCOP2, aims to define and annotate relationships such as internal structural repeats, common motifs and subfolds that have not been a subject of classification in the SCOP database. Currently, these non-hierarchical relationships are subdivided into five different types.

Biounitgen refers to a relationship that occurs between and within the biological units. Gene duplication/fusion and other evolutionary events in the proteins that form oligomeric biological units often result in monomeric multidomain proteins, which structures resemble these oligomeric biological units. The Biounitgen relationship is represented by a structural region shared by the subunits of the former and the domains of the latter. Example: hydroxyisobutirate dehydrogenase dimerisation/6-phosphogluconate dehydrogenase duplication region.

Foldogen defines a relationship between homologous structural domains that share a large common subfold but differ in the regions that complement it to a globular fold. This relationship is introduced mainly for the annotation of proteins and protein families known to adopt more than one ‘native’ fold. Example: Common subfold of the different types of the cro folds.

Motif is a local structural region of common sequence and/or function shared by both homologous and non-homologous proteins. Example is the Helix-Turn-Helix (HTH) motif.

Repeated domain refers to large, domain-sized structural repeats with a low sequence similarity, analogous to the intermolecular ‘superfamily’ relationships. (In general, the multi-copy domains of similar sequences are classified into their respective families.) Unlike structural domains of the same protein, the repeated domains may swap with their equivalent segments and therefore be non-compact. Example: D-2-hydroxyacid dehydrohenase-like half-domain.

Repeated motif defines small structural units of similar sequences that are building blocks of  larger proteins, for example, Ankyrin repeat.

1.2.2. SCOP2 domains.

In SCOP2, there is no a priori division of protein structures into domains, in which one domain size fits all possible relationships. The protein domain is defined as a unit of relationship and its boundaries are dependent on a given relationship. In general, the protein domain, representing a child node of the SCOP2 graph, is larger than that representing its parental node domain of the same protein, especially if there is more than one parental node for a given child node. For example, Fold is an attribute of a single structural domain, but the domains representing Family and Superfamily can span over more than one structural domain. Similarly, the Family domain may contain more than one non-overlapping Superfamily domains.  Example: Tm_MurC.

1.3. SCOP2 representative sequences and structures.

The SCOP2 classification is based on the representative sequences (UniProt) and structures (PDB).  The manual annotation of these representatives is then automatically extended to related entries.

Only those structures with a complete or near complete coverage of the full-length protein chains are represented at the Species level. The structures of fragments corresponding to one or more family domains are classified at the corresponding levels. The isolated family-sized domains are grouped together at the family node, whereas the structures of multidomain fragments and structures of different individual domains from the same protein can be grouped at the protein level. To distinguish them from the full-length structures, the structures of protein fragments are annotated with appropriate keywords.

1.4. SCOP2 controlled vocabulary ( keywords and tags ).

Additional annotation of the SCOP2 entries is provided using a controlled vocabulary (keywords and tags). This allows easy retrieval of a subset of proteins with a given feature or automatic processing of the classification data. Example: fragment (multi-family coverage).

2. Getting started. Browsing and searching SCOP2.

The SCOP2 data can be accessed, using the main navigation menu, in two different ways: via SCOP2-browser and SCOP2-graph. SCOP2-browser allows navigation through the SCOP2 classification in a traditional way by browsing pages displaying the node information. An alternative way of viewing the SCOP2 data is by using the SCOP2-graph. SCOP2-graph is a web-based viewer for display and navigation through the graph of SCOP2 nodes.

2.1. SCOP2-graph.

SCOP2-graph consists of three key components: Main window, Node popup window and toolbar. The main window in the centre displays the ancestry tree for a selected node. The node popup window displays additional information about the node and appears when you click on a node. The toolbar with action buttons is located at the top and allows to configure the display (Configure button), share the view with a colleague (Share button) and return to the default display (Reset button).

When you click on the Help button, a popup window appears with information on currently active component. You can keep the help popup open while using the SCOP2-graph - the help info will be updated accordingly as soon as you click on a different component.                                            

2.1.1. SCOP2-graph Main window.

SCOP2-graph is a graph-based tool for display and navigation through the graph  of SCOP2 nodes. By default  SCOP2-graph displays the top level of the SCOP2 classification. You can click on the displayed nodes to bring the Node popup window. You can expand or collapse the displayed graph  by selecting  given parents and/or children of interest in the Node window. You can also  see additional information about the node, e.g representative sequence and structural regions, comments, structural tags  etc.  

By default the SCOP2-graph displays all the nodes and relations in colour. The colour legend is displayed in the collapsible panel on the left . Similarly, by default the graph shows all ancestors of the selected node and does not include the node descendants. You can change the default display via Configuration panel if you click on Configure button in the toolbar at the top.  

If you want to share the tree you see with a colleague you just need to click on Share button and copy  the displayed URL.

If you would like to start a new session you can use the Reset button to restore the default display.  The Reset button will clear the session data and will display only the nodes at the top level of the SCOP2 classification..    

2.1.2. SCOP2-graph Node popup window.

The window allows you to navigate through the graph and to display the node details. There are four action buttons at the top of the Node popup window:

Focus : refocus and displays the ancestry tree for a given node.

Remove : removes the node from the tree as well as all node descendants which do not have other ancestors except for this node.

Browser : displays the node in the SCOP2 browser.

SCOP: provides a link to the node in SCOP from which the SCOP2 node originates.

The Node popup window contains four sections (tabs), namely Structure, Parents, Children and Comments.

2.1.2.1. Structure tab.

The Structure tab provides visualization of the representative of the given node. The region of protein 3D structure representing this node is displayed using the Jmol application. By default, the structure of the entire pdb entry is displayed and the representative region is coloured by structure. You can change the default display of protein structure via the Configure button located at the top toolbar.  

The full length representative sequence is shown at the bottom of the Structure window and the representative sequence region for this node below it. The Structure window also provides hyperlinks to view the representative structure and sequence in external databases such as PDB, ePDB and UniProt.

If you select a domain from the Domains dropdown listyou will see which parts of the protein constitute the representative domain. The segments field defines these parts with PDB residue numbers and chain ID. The Jmol applet shows where these parts are in the protein structure and the representative sequence diagram at the bottom shows where they are located in the sequence. You can click on a protein bar, or a segment bar in the reference sequence diagram to retrieve the corresponding sequence.

2.1.2.2. Parents and Children tabs.

The Parents and Children tabs contain immediate ancestors and descendants of the node. The parent and child nodes are coloured according to the type of the relation between them and the node of interest. You can select which children and/or parents should be added and/or removed from the display.

User can select all nodes or just a selection from those that are listed and click on update in order to display them on the graph. There are some selection limitations due to the nature of the classification graph. Only up to 10 children of any given node can be displayed at the same time.

2.1.2.3. Comment tab.

The Comment tab displays some classification comments made for clarity and controlled vocabulary terms associated with the node.

2.1.3. SCOP2-graph Configuration window.

The available configuration options are as follows :

Show relations : allows you to specify which relations should be included in the graph. By default all relations are displayed and all node ancestors are shown. If you un-check a relation type then all relations of this type will be skipped from the graph as well as all the nodes which have been connected by them.

Domain display : configures the default cartoon type used to display protein structures in the Jmol applet located in the Structure tab of the Node popup window.

On node focus : defines what should happen to the selected node children. By default the children are not shown, i.e "collapsed".

Classes display : by default the nodes and relations are coloured according to their type. Single-colour option will force everything to be displayed in the same colour.

2.1.4. How to search SCOP2-graph.

The SCOP2-graph search engine allows queries by general text, PDB ID, UniProt ID, SCOP2 ID and keywords from the controlled vocabulary. To perform a search use the search box at the top right corner of the browser window. Focus the cursor into  the box and start typing a search term. When four characters are entered the autocomplete dropdown list will appear populated with SCOP2 terms satisfying the search criteria. If you can see the SCOP2 term you are interested in you can just click on it. Once you have selected a node its ancestry tree will be displayed and the selected node will be highlighted with a light blue background. If you want to go back to the search results navigate the cursor back into the search box and press ‘arrow down’ button - it will bring back the results of the last search. It is also possible to perform search of terms that contain two or more words. These have to be concatenated by % sign, e.g.  binding%domain. The search is case-insensitive, but the order of the words is important.

If you want to see the search results again, make sure the focus is in the search box and simply press "cursor down" key to bring up the search results popup window again.

2.2. SCOP2-browser.

SCOP2-Browser allows accessing the database content in a more traditional way by browsing the pages displaying the node information. The data are accessed via a drop down menu which provides an entry point for the main classification categories. Typically, the node page displays the node of interest boxed in gray. At the edge of this gray box a hyperlink allows to locate the node on the SCOP2-graph. Above the node of interest are listed the parental nodes and below the children. At the bottom of the node page a table lists the domain information. The representative sequence and structural domains are listed on the left. Their boundaries are defined manually. Links to external databases such as UniProt and PDB (click on the icons) are provided as well as a link to the SCOP domain from which the SCOP2 representative domain originates (if it exists). In addition, hyperlinks provide access to the representative domain page and the Cross-reference map (click on domain boundaries). These are explained in more detail below. The right section of the table lists the automatically assigned cluster domains.

2.1.1. Representative domain page.

The domain page lists the representative domain information. A Jmol applet shows the structure of the representative region. Example: Representative domain for HypF zinc finger-like fold.

2.1.2. Cross-Reference Map.

The cross-reference map is a drawing showing the representative sequence at the top, followed by cartoons of PDB chains and SCOP2 domains that correspond to a part of or the whole of the representative sequence. Light green cartoons represent domain sequences in SCOP2. Clicking on a cartoon will take you to the SCOP2 page for that domain. Note that each segment in a multi-segment domain is shown as a separate bar. Blue cartoons represent segments of PDB chains that are mapped to the representative sequence by SIFTS. Clicking on a cartoon will take you to the RCSB page for that PDB entry. Note that a single PDB chain may show as several segment bars if SIFTS does not map the entire chain. Example: Cross-reference map for representative domain for HypF zinc finger-like fold.

2.1.3. Clustered Domains.

Clustered Domains are comprised of segments of those PDB chains that SIFTS cross-references to a SCOP2 domain via the domain's representative sequence from UniProt. So they are very similar to the domains defined manually in SCOP2. Their boundaries are derived automatically without human supervision and thus reliability on their assignment depends on the quality of the cross-reference data.

Some clustered domains have segments shortened because the residues at the end of the segment were not present or were unobserved. Such domains are highlighted in blue. Check the boundaries of such domains carefully to be sure they are appropriate for your use.

Other clustered domains are highlighted in red. These domains have unobserved residues in the interior part of their segments. They may also have segments shortened because the residues at the end of the segment were not present or were unobserved. Check the details of such domains carefully using original source data, to be sure they are appropriate for your use.

2.1.4. How to search SCOP2-Browser.

To find pages in the browser by content, use the Search box located at the top right corner. The search engine allows search by PDB ID, UniProt accession, general  text that is the complete value of a field, such as "GroES-like" or ID number of a  SCOP2 node. If you want to search for a word or phrase that occurs somewhere in a field, but which is not the whole field, then add a '*' to the end of the word. If the search finds a single page containing the search term, it will redirect you to that page. If it finds several pages, it will show a list of them to allow you to browse one or more of them.

3. Molecular viewers.

3.1. Jmol.

Both navigation tools, SCOP2-Browser and SCOP2-Graph, visualize the representative structures of SCOP2 nodes using the Jmol applet (an open-source Java viewer for chemical structures in 3D, http://www.jmol.org/). Some basic commands are listed below, for more information refer to the official Jmol site.

3.1.1. Basic commands.

 

Open Jmol Menu: Right Click ; Ctrl + Left Click ;

Rotate Around X,Y : Left Click and Drag

Move along X,Y (= translate) : Shift + Left Double-Click and Drag Middle Double-Click and Drag ; Ctrl + Right Click and Drag

* Works both when clicking on the molecule or away from it.

Reset and Center : Shift + Left Double-Click ; Middle Double-Click

* Only works if double-click is done away from the molecule.

Rotate Around Z : Shift + Left Click and Drag Horizontally ; Middle Click and Drag Horizontally ; Shift + Right Click and Drag Horizontally (possibly fails on OS X)

Zoom In / Out : Shift + Left Click and Drag Vertically ; Middle Click and Drag Vertically

Use Mouse Wheel    

 

3.2. Rasmol and rmscop configuration.

SCOP2 browser provides an interface to RasMol via the rmscop script for display of protein structures. With the rmscop script correctly installed, clicking on a rasmol icon in SCOP2 browser will cause a local copy of RasMol to be started, displaying the molecule selected. The latest RasMol version and help with the viewer can be downloaded from http://rasmol.org/  and the rmscop from here.

3.2.1. Installing Rmscop.

This page contains technical information for unix/linux users only. A general overview about using RasMol to view structures in SCOP2 can be found here.

It is assumed that you have both tcl/tk and RasMol correctly installed on your system. To install and configure rmscop on your computer, follow the instructions. If you have any problem, don't hesitate to contact us.

* Download rmscop and save it in a directory in your path (typically, ~/bin).

* Edit rmscop to configure it. Search for SITE-CONFIG. There are 4 sections:

1. set pdbdir {</pdb_disk/pdb_directory/>} where </pdb_disk/pdb_directory/> is the full pathname to your local pdb directory (if you have one) Use: set pdbdir {} if you don't have local pdb files set pdbpref {<pdb prefix>} where <pdb prefix> is the pdb filename prefix (i.e. pdb in pdb2cro.ent). It can be empty ({}) set pdbext {<pdb extension>} where <pdb extension> is the pdb filename extension (i.e. .ent in pdb2cro.ent). It can include a compression suffix ({.ent.Z})

2. set rasmol {<rasmol name>} <rasmol name> must be the name of your RasMol executable file (usually, rasmol)

3. set ftp parameters (ftpserver, ftpserverdir, ftpext, cmpext)

4. set ftpdir {<tmpdir>} where <tmpdir> specifies the full path to an area on your system with write permission. Files that are not found at </pdb_disk/pdb_directory/> will be downloaded and stored at <tmpdir>

* Change file protections (chmod 755 rmscop)

* Add the following line to your .mailcap file: application/x-rasmol; rmscop %s (this file should be in your home directory. If it does not exist, simply create it) You may need to source your .cshrc, .login or equivalent file, and start your web browser again before changes become effective. Clicking on a rasmol icon () from a SCOP2 classification page should open a RasMol window and show the selected molecule.

If you want to have two structures displayed at the same time, launch (from the shell) rmscop twice. This will open two RasMol windows. Then click on two different rasmol icons (). You should get a window that allows you to choose in which of the two RasMol windows the molecule would be displayed

4. Linking to SCOP2.

Linking to pages in the SCOP2 browser

To link to the page for any node just use the node identifier. So examples for a fold and a domain would be:

       http://scop2.mrc-lmb.cam.ac.uk/2000131

       http://scop2.mrc-lmb.cam.ac.uk/8003965

For other types of pages there is a prefix as well. For example, the prefix for representative sequences is 'rep-', so the sequence page linked to the above examples is:

       http://scop2.mrc-lmb.cam.ac.uk/rep-10159

For other types of pages, just look at any page of that type to discover the appropriate prefix. You can also add '.html' as a suffix if you wish.

5. SCOP2 identifiers.

Each node in SCOP2 can be uniquely identified by a seven digit identifier. The SCOP2 node identifiers will be stable across releases but may become obsolete if significant changes are introduced to the node.

The SCOP2 representative domains are also represented by domain identifier e.g. CF-8004435-3VTHA , that allows easily to locate the level that domain represents and the PDB and chain ID(s) of the representative structure from which it is derived.

6. Download.

SCOP2 data can be downloaded here. A README file describes details about the content of files and release changes.

7. SCOP2 REST web-service.

The SCOP2 REST web-service is a convenient way to retrieve data from SCOP2. It allows easy programming language-agnostic access to SCOP2 ontology and SCOP2 domains. The data can be requested with simple HTTP requests and returned in JSON. A variety of programatic and bioinformatics relevant formats may be  supported in the future.

7.1.  Making a REST request.

Each datatype resides on a REST endpoint, these are all listed in the Endpoints section. A request is made to an endpoint as a parameterised HTTP request adhering to a predefined URI schema. Parameters can be required or optional. Optional parameters are indicated by being placed inside square brackets, e.g [ p = 1], in a documented endpoint. Other parameters are required and must be included in the URL. All parameters can be included as CGI style parameters as part of the HTTP request.

7.2. Example request.

for node - provides all the info about the SCOP2 node, including its parents, children, domains etc, e.g                                              
http://scop2.mrc-lmb.cam.ac.uk/graph/restapi/term?id=SF:3000133

for domain - provides the info about the SCOP2 domain, including its segments data and the SCOP2 node, e.g                                                                                  
http://scop2.mrc-lmb.cam.ac.uk/graph/restapi/domain?id=CF-8004382-2BCGG

7.3. Content Type.

At the moment only JSON format is supported, but we'll add others depending on user requests.

More information about endpoint and access can be found here.

8. Acknowledgements.

 The SCOP2 authors wish to acknowledge Marco Punta, Andreas Prlic, Spencer Bliven for their helpful comments and suggestions.