RING - Residue Interaction Network Generator

Graph theory has been applied to the structure of macromolecules and proteins in particular for some time. A (likely incomplete) list of relevant literature can be found here. Proteins can be analysed using graph theory through residue-residue "Interaction Networks". The analysis of the network of interactions between the amino acids of a protein may be useful to derive new knowledge regarding the significance of various network parameters.

RING (Residue Interaction Network Generator) is a web server for transforming a protein structure into a network of interactions. Nodes represent single amino acids in the protein structure, while the edges represent the non-covalent interactions that exist between them.

RING is able to process a single PDB file and output a number of files encoding the interaction network. The network is built on the protein chain and its attributes. The created network and its edge attributes are stored in files using the Simple Interaction Format (SIF). These files can then be easily loaded in CYTOSCAPE to visualize and manipulate the network.

The "degree" of a node means the total number of connections (and therefore the number of neighbors) that it possesses. Amino acids showing high connectivity in the network generally play an important role in the structure/function of the protein. RING therefore produces a multiple sequence alignment, in FASTA format, where each alignment is filtered according to the increasing degree of corresponding nodes in the structure.

In cases where the three-dimensional structure of a protein is known it can be assumed that there is a physical interaction between two amino acids if they are close enough to each other. RING defines the interaction between a pair of residues (with a minimum sequence separation) in two main ways:

• as the threshold distance between the atoms composing them. Comparisons are either made according to the threshold distance between C-alpha carbons or between the closest atoms of the residue pair.


• if there is at least one van der Waals interaction between pairs of atoms that compose them respectively.

• Simple interactions
  • Pairs of C-alpha atoms
  • Pairs of closest atoms

• van der Waals interactions

• Hydrogen bond

• Salt bridge

• π-cation interactions

• π-π interactions

• Disulfide bridges

• Peptide bonds

The classification of interactions between a pair of residues is mutually exclusive. For example, it is clear that in nature hydrogen bonds can exist either alone or in combination with one of the other classified interaction types but RING assumes they are distinct. RING does not take into account the presence of any hydrogen bonds that coexist with other interaction outside the van der Waal types.

RING also specifies which portion of an amino acid (side chain or main chain atoms) is involved in a given interaction. Furthermore, the network connections identify certain electrostatic interaction types where partners have very different charges (such as hydrogen bonds, salt bridges and π-cation interactions). These are directed positive-charge → negative-charge, when displayed in CYTOSCAPE.

RING first of all submits the PDB file to be analyzed by the program REDUCE. This positions amide groups correctly in the Asn, Gln side chains, the methyl group of Met, the imidazole ring of His while also optimizing the orientation of many other polar groups. It maximizes the potential hydrogen interactions between residues. REDUCE also accurately determines the coordinates of missing hydrogen atoms, according to the optimized orientations (see reference).

The basic connectivity of the network is calculated according to the chosen type:

• Van der Waals interactions;
• Closest atom distances;
• C-alpha distances.

Van der Waals interactions are calculated by the program PROBE witch identifies the contacts between amino acids in a protein by evaluating their atomic packing using small-probe contact dot surfaces (see reference).

Once the interacting pairs are defined, each interaction is classified in physico-chemical terms and the multiple sequence alignment based on the increasing degree of matching nodes is calculated in parallel, starting with the threshold.

Given the importance and complexity of the nature of hydrogen bonds, RING interfaces with the integrated HBEXPLORE program for their detection and allows the user to choose one of the search criteria made available by the program (or a combination of them. See reference).

At the end of execution RING prepares an archive containing all files describing the network, their edge and node attributes, the multiple sequence alignment based on the nodes degrees and their analysis.

Interfaces

RING offers to the user two interfaces:

• A simple interface where the user is prompted to insert a minimal number of options. These are needed for the program functions while default parameters are used.

• For more demanding users, options can be set in a finer way. These determine all the geometric parameters used to classify the interactions and to guide the function of the integrated programs.

An optional title for your submission. This will appear in the header of the output. We suggest you select one in order to better identify your job, especially in case of multiple submissions.

A PDB file of the structure from which to derive the network is required. This PDB file can either be selected from the local PDB database by providing a valid ID, or uploaded using the appropriate button. The PDB file may contain more than one chain. In this case, the chain identifier has to be supplied, otherwise the first chain will be selected.

NB: Please make sure that the Chain selection is either left blank or filled with a valid identifier for the server to produce meaningful results.

RING is able to generate residue interaction networks from a protein structure in various ways by combining two sets of information:

• The general connectivity between residues based on van der Waals contacts or distance threshold between atoms. This information is used to define the architecture of the network.

• The physical/chemical interaction properties of the identified types (in the case of hydrogen bonds and van der Waals interactions it may provide further connectivity data). This information defines the attributes of edges in the network.

RING is capable of generating three different types of interaction network to be displayed in CYTOSCAPE. Each type provides a different set of edge attributes. The three types are:

• a more realistic network based on van der Waals contacts.

• a complex network based on distances between the closest atoms.

• a simple network based on distances between C&alpha atoms (complex interface only).

To build the network based on van der Waals interactions, RING relies on PROBE a program developed to analyze atomic packing whose approach is to place a small probe (of radius 0.25 Å) at points along the van der Waals surface of a selected set of atoms and determine if this probe also contacts atoms within a second "target" set. Between outputs it can produce a detailed description of each dot, including source and target atom names and partial scores.
RING uses this to calculate his van der Waals network (see method).

Minimum distances   (complex interface only)

Minimum interaction distance

The form records the distance threshold on which an interaction is defined. It threshold works on the chosen type of network:

• C-alpha network represents the minimum distance required between C_α residue pairs in order to draw an edge;

• Closest atoms network represents the minimum distance between the closest atoms of a pair of residues in order to draw an edge.

Minimum disulfide bridge distance

A disulfide bridge is a a covalent bond between 2 atoms, which presents a more or less constant distance. RING uses this type of constraint to identify a disulfide bridge between two Cys residues in contact. They are considered covalently bound if the distance between sulfur atoms (S_γ) is less than the threshold value.

Unless specified by the user, a 3 Å default threshold is used. This was empirically derived from the average length of the covalent bond. For more information please see method details.

Minimum salt bridge distance

RING determines the presence of a salt bridge at physiological pH when a negatively charged residue, i.e. Asp or Glu, is in contact with a positively charged residue, i.e. Arg, Lys or His. These types of amino acids are considered involved in a salt bridge if the distance between the mass centers of the charged groups in their side chains is less that the distance threshold, in Angstrom (Å).

Unless set by the user, a 4 Å default threshold is used. This threshold was empirically derived from analysis of a large set of protein structures. For more information please see method details.

Minimum π - π distance

Given a pair of residues defined in contact, RING checks if both are aromatics, i.e His, Tyr, Trp or Phe. After that, a π-π interaction exists between them if there is at least one pair of atoms, one for each of the two side chains, at a distance less than the threshold value.

The default threshold value is 6 Å, derived empirically from analysis of a large group of protein structures. For more information please see method details.

Minimum π - cation distance

RING searches for π-cation interactions when two residues in contact are: a positively charged amino acid at physiological pH, i.e. Arg or Lys; and an aromatic amino acid, i.e. Phe, Tyr or Trp. Histidine, while possessing an aromatic side chain is not considered, because it can participate in this interaction type both as a cation and as a π-system (depending on its protonation state). For π-cation pairs, the main condition to be met is that the mass center of the charged group in the side chain of a cation should be located at a distance less than the threshold value, in relation to any atom of the other π-system residues.

The empirically derived default threshold value is 7 Å. For more information please see method details.

Minimum sequence position distance

RING aims to highlight only interactions between residues close in three-dimensional structure of the protein, but sufficiently distant in sequence. Minimum Sequence positions allow control over the minimum sequence distance. If the user does not specify any value, the default is a minimum sequence separation of 2 positions.

Options   (complex interface only)

Hydrogen atom options

RING relies on the external program REDUCE to add hydrogen atom coordinates to molecular structure files. REDUCE also corrects orientation errors of important chemical groups in the side chain of certain amino acids such as Asn, Gln, Met and His.

REDUCE first assigns the correct orientation of amide groups in Asn and Gln side chains. It optimizes the orientation of hydroxyl (-OH), sulphydrate (-SH), amino (-NH₃⁺), methyl in Met groups and correctly orients the imidazole ring in His. REDUCE then creates a new PDB file with the coordinates of the missing hydrogen atoms consistent with the newly assigned amino acid positions. This PDB file is made available to the user on the results page (for more details see output page).

RING offers three options regarding the execution of REDUCE:

• The default option is Replace all. This option removes any of hydrogen atom coordinates already present before creating the new positions.

• Alternatively, the option Keep existing forces the program to keep hydrogen atoms already present in the PDB file (e.g. experimentally resolved) and generate only the missing ones.

• Finally the option No correction bypasses the hydrogen atom correction execution and takes directly the original PDB file for subsequent analysis with the hydrogen atom coordinates added directly by HBexplore.

For details about the operation by REDUCE go to the page method details.

Hydrogen bond policy

Given the importance and complex nature of hydrogen bonds, RING commits their identification to the external program HBexplore.

RING allows the user to choose one of two search criteria provided by HBexplore or to choose a combination of the two criteria:

• the user can choose whether to adopt the first criterion;

• the second more reliable criterion takes into account the geometry of the free orbital of the acceptor;

• the third criterion combines the previous two, i.e. it imports hydrogen bonds accepted by the second criterion plus those recognized by the first but not the second.

By default the second criterion is selected. For more information please see method details.

Node degree sequence alignment

RING implements a feature that produces a file containing a multiple sequence alignment in FASTA format. Built on the amino acid sequence of the protein under consideration, each line contains the sequence filtered with increasing node degrees. If the checkbox Node degree sequence alignment is selected, the multiple sequence alignment is constructed starting from the defined degree threshold in the text form Starting threshold up to the maximum node. The default starting value is 1.

In addition to the multiple sequence alignment, RING produces a text file with extension .anal containing useful analyses of the node degrees for the protein residues. For more information please see method details.

Sub-Networks and other Node Features   (complex interface only)

Conservation in multiple sequence alignment

RING implements a feature that produces a file containing a multiple sequence alignment in FASTA format. This file is produced by running PSI-BLAST against the UniRef90 sequence database. In the Complex interface it is possible to specify the number of iterations an E-value of the sequences included in it, otherwise defoult paremeters are one iteration and E-value threshold of 0.001.

In addition RING, in the complex interface only, allows to produce sub-networks containing only residues with a conservation threshold above a user defined conservation threshold.

For more information please see method details and PSI-BLAST reference.

(Pairwise) Mutual Information (MI)

RING calculates the ammount of mutual information between pairs of residues positions in the sequence alignment generated to compute conservation. MI, comulatice MI and its correction APC for those amino acids present in the final network.

For more information please see method details and MI reference.

Solvent Accessibility and Secondary Structure assignments

RING also produces files containing solvent accessibility and secondary structure for each of the amino acids present in the submitted structure. Solvent accessibility and secondary structure are assigned by the program DSSP. Solvent accessibility is normalised by the observed solvent accessibility for GLY-Xaa-GLY in fully extended conformation, values were taken from Miller et al.. A secondary structure assignment based solely on ϕ and ψ torsion angles is also generated in RING.

Energy Calculation   (complex interface only)

FRST Energy Calculation

RING provides a file containing FRST residue based energy calculation.

For more information please see method details and FRST reference.

TAP Score Energy Calculation

RING provides a file containing TAP residue based energy calculation.

For more information please see method details and TAP reference.

Peptidic Bonds   (complex interface only)

Unless specified otherwise by the user RING provides an adds edges to the main network file representing peptidic bonds between the amino acids present in the network.

Output

RING generates an interaction network to be loaded in CYTOSCAPE with different interaction types.
A detailed description of the RING output can be found on the output page.

Examples

Below are the links to the output for several different examples. A description of these can be found in the tutorial .

• Example 1 - Closest atom network from crystal structure of human glutathione peroxidase 4 (GPX4), at 1.55 Å resolution (PDB code: 2OBI).

  • Chain A.
  • Parameters:
    • Correction with H removal of hydrogen atoms.
    • Second criterion hydrogen bond policy.
    • Node degree sequence alignment.
    • Conservation multiple sequence alignment.
    • Solvent accessibility and Secondary structure.
    • FRST and TAP Score energy evaluation.
  • The corresponding Cytoscape session can be downloaded here.

• Example 2 - Closest atom conserved residues (80%) sub-network from crystal structure of human glutathione peroxidase 4 (GPX4), at 1.55 Å resolution (PDB code: 2OBI).

  • Chain A.
  • Standard parameters (see example 1 above).
  • The corresponding Cytoscape session can be downloaded here.

• Example 3 - Closest atom network of the Rhamnose-binding lectin csl3 at 1.8 Å resolution (PDB code: 2ZX2).

  • Chain A.
  • Standard parameters (see example 1 above).
  • The corresponding Cytoscape session can be downloaded here.

• Example 4 - Closest atom network of human triosephosphate isomerase (TIM barrel) at 2.8 Å resolution (PDB code: 1HTI).

  • Chain A.
  • Standard parameters (see example 1 above).
  • The corresponding Cytoscape session can be downloaded here.

• Example 5 - Closest atom network of ABL tyrosine kinase SH3 domain with 3BP-1 synthetic peptide at 2.0 Å resolution (PDB code: 1ABO).

  • Chain A.
  • Standard parameters (see example 1 above).
  • The corresponding Cytoscape session can be downloaded here.

• Example 6 - Closest atom network of Ectodomain of human ADAM22 at 2.36 Å resolution (PDB code: 3G5C).

  • Chain A.
  • Standard parameters (see example 1 above).
  • The corresponding Cytoscape session can be downloaded here.

References

If you use this server in a work leading to publications, please cite:

• RING web server:
  Alberto J.M. Martin, Michele Vidotto, Filippo Boscariol, Tomás Di Domenico, Ian Walsh and Silvio C.E. Tosatto.
  RING: Networking interacting residues, evolutionary information and energetics in protein structures.
  Bioinformatics,   2011 Apr 14. [Epub ahead of print].   (2011)

A (likely incomplete) list of relevant literature can be found here. Additional references of the components used by the server are:

• REDUCE:
  J. M. Word, S.C. Lovell, J.S. Richardson and D.C. Richardson.
  Asparagine and Glutamine: Using Hydrogen Atom Contacts in the Choice of Side-chain Amide Orientation.
  J. Mol. Biol. 285, 1735-1747   (1999).

• PROBE:
  J. M. Word, S.C. Lovell, T. H. LaBean, H. C. Taylor, M. E. Zalis, B. K. Presley, J. S. Richardson and D. C. Richardson.
  Visualizing and Quantifying Molecular Goodness-of-Fit: Small-probe Contact Dots with Explicit Hydrogen Atoms.
  J. Mol. Biol. 285, 1711-1733   (1999).

• HBexplore:
  K. Lindauer, C. Bendic and J. Sühnel.
  HBexplore - A new tool for identifying and analyzing hydrogen bonding patterns in biological macromolecules.
  Comput. Appl. Biosci. 12, 281-289   (1996).

• DSSP:
  W. Kabsch and C. Sander.
  Dictionary of protein secondary structure: pattern recognition of hydrogen-bonded and geometrical features.
  Biopolymers. 22(12), 2577-2637   (1983).

• Solvent accessibility:
  S. Miller, J. Janin, A. M. Lesk and C. Chothia.
  Interior and surface of monomeric proteins.
  Journal of Molecular Biology. 196(3), 641-656   (1987).

• PSI-BLAST:
  S.F. Altschul, T.L. Madden, A.A. Schaffer, J. Zhang, Z. Zhang, W. Miller, and D.J. Lipman.
  Gapped blast and psi-blast: A new generation of protein database search programs.
  Nucleic Acids Research. 25(17), 3389-3402   (1997).

• FRST:
  S.C.E. Tosatto.
  The Victor/FRST Function for Model Quality Estimation.
  Journal of Computational Biology. 12(10), 1316-1327   (2005).

• TAP:
  S.C.E. Tosatto and R. Battistutta.
  TAP score: torsion angle propensity normalization applied to local protein structure evaluation.
  BMC Bioinformatics. 8, 155   (2007).

• Mutual Information:
  C.M. Buslje, J. Santos, J.M. Delfino and M. Nielsen
  Correction for phylogeny, small number of observations and data redundancy improves the identification of coevolving amino acid pairs using mutual information.
  Bioinformatics. 25, 1125-1131   (2009).