Nucleic Acids Research Advance Access originally published online on October 30, 2008
Nucleic Acids Research 2009 37(Database issue):D224-D228; doi:10.1093/nar/gkn833
Nucleic Acids Research, 2009, Vol. 37, Database issue D224-D228
© 2008 The Author(s)
This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/by-nc/2.0/uk/) which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited.
Protein segment finder: an online search engine for segment motifs in the PDB
Abraham O. Samson* and
Michael Levitt
Department of Structural Biology, Stanford University, Stanford, CA 94305, USA
*To whom correspondence should be addressed. Tel: +1 650 725 0754; Fax: +1 650 723 8464; Email: avraham.samson{at}stanford.edu
Received August 15, 2008. Revised October 10, 2008. Accepted October 14, 2008.
 |
ABSTRACT
|
|---|
Finding related conformations in the Protein Data Bank (PDB)
is essential in many areas of bioscience. To assist this task,
we designed a search engine that uses a compact database to
quickly identify protein segments obeying a set of primary,
secondary and tertiary structure constraints. The database contains
information such as amino acid sequence, secondary structure,
disulfide bonds, hydrogen bonds and atoms in contact as calculated
from all protein structures in the PDB. The search engine parses
the database and returns hits that match the queried parameters.
The conformation search engine, which is notable for its high
speed and interactive feedback, is expected to assist scientists
in discovering conformation homologs and predicting protein
structure. The engine is publicly available at
http://ari.stanford.edu/psf and it will also be used in-house in an automatic mode aimed
at discovering new protein motifs.
|
INTRODUCTION
|
|---|
Over the past years, structural data in the Protein Data Bank
(PDB) has grown enormously (
1). This is due both to the worldwide
structural genomics effort, and to recent advances in X-ray
crystallography, such as high-intensity synchrotron beam lines.
At the present time, approximately 52 000 protein and nucleic
acid structures are available (release of August 2008). So much
data makes it difficult to navigate the information and realize
its wealth, particularly when one tries to retrieve segment
motifs (
2). Retrieval of such segment motifs is often used in
widespread applications such as protein structure prediction,
loop modeling and homology matching.
To assist this task and sort quickly and efficiently through available structures, computational engines providing fast structural queries of the PDB have been designed. One such basic engine is the advanced search option of the PDB which handles queries for amino acid sequence and secondary structure amount (3). A more advanced program is MSDMOTIF which can combine searches for sequence motifs, structure motifs, protein sequence, 3D properties secondary structure elements, etc. (4). Although very reliable, the program is resource-intensive and response is very slow limiting its use. Another advanced program is SPASM which was developed by Kleywegt and coworkers (5) to find spatial motifs consisting of arbitrary main-chain and side-chains conformation in a database of protein structures. This program is fast and excels in finding spatial homologs displaying low RMS deviations for a set of PDB coordinates. An additional search engine named Fragment Finder was designed by Sekar and coworkers (6) to identify similar 3D structural motifs. This program is based on the similarity of backbone 
and 
dihedral-angles and allows superimposed display of search results. Another engine, PAST, which was developed by Griebsch and coworkers (7), is based on translation- and rotation-invariant representation of protein backbone. Takahashi and coworkers (8) developed a 3D substructure search program named SS3D-P2 to find protein motifs based on secondary structure elements. Akutsu et al. (9) developed another engine to rapidly search for protein segment homology based on Fourier transforms. In this engine, the similarity of segments is evaluated from the difference between hash vectors consisting of low frequency components of Fourier-like spectrum for the distances between the C
atom and the centroid. Last but not least, Helmer-Citterich et al. designed a search engine named PdbFun for structural and functional analysis of proteins at the residue level. PdbFun executes searches using various criteria such as secondary structure, residue type, protein domain, solvent exposure, ligand binding ability and catalytic activity (10). The aforementioned engines were all designed to identify either spatial similarity or sequence similarity but do not allow combined search for segment motifs using primary, secondary and tertiary structure constraints. There is a clear and so far unmet need for a fast search engine, which combines these query constraints and includes amino acid sequence, secondary structure and contacts.
Here, we describe an online search engine that rapidly finds peptide segments which satisfy a set of conformational parameters in available structural data of the PDB. Query parameters include amino acid sequence, sequence motifs, secondary structure, secondary structure elements, disulfide bonds, hydrogen bonds and residue contacts. Public access to the search engine is facilitated through a simple interactive graphic user interface (GUI). The search engine, named Protein Segment Finder (PSF), is advantageous due to its speed, generality and simplicity. It is expected to be helpful to the scientific community by easing the identification of segment motifs (2) and conformation homologs, and distilling useful information from the PDB.
|
DATABASE
|
|---|
To avoid the time-consuming task of calculating protein contacts
for every search, a database containing contact maps of all
PDB proteins was prepared. The database contains entries for
all protein structures in the PDB, in the format shown in
Figure 1.
Each database entry is headed by a line consisting of the PDB
ID and the protein name separated by a vertical bar. This header
is followed by two lines, one with the amino acid sequence and
another with the secondary structure both in the one letter
code. The amino acid sequence is all in capital letters, except
for cystine pairs which are denoted by small letter pairs. The
secondary structure sequence, corresponding to the sequence
of amino acids with a particular secondary structure, was calculated
using the DSSP program by Kabsch and Sander (
11) in which H
represents
-helix, E extended β-structure, B isolated β-structure,
T hydrogen bonded turn, S bend, I
-helix, G 3
10-helix and spaces
random coil.
View larger version (35K):
[in this window]
[in a new window]
[Download PowerPoint slide]
|
Figure 1. Sample database entry and structure for a typical small protein. Shown is the database entry for PDB ID 1em7 together with the ribbon diagram of the same protein (in inset). Line 1 is the sequence, line 2 is the secondary structure and lines 3–32 indicate contacts between position 1 and a position up to 30 residues further along the chain.
|
|
Next comes the contact map which is currently limited to a size
of 30 lines for reasons of speed. This contact map summarizes
all protein contacts separated by up to 30 residues. Each line
of the contact map begins with a dot followed by an index number
composed of two digits. The two digit index number corresponds
to the number of amino acids separating all interactions on
that line. Contact types are recorded using a single code digit
ranging from ‘1’ to ‘8’ that denotes
the occurrence of heavy atoms and C
interactions as well as
hydrogen bonds. These code digits are defined as follows: ‘1’
indicates contact between heavy atoms; ‘2’ indicates
a contact between C
pairs; ‘3’ indicates a backbone
HN · CO hydrogen bond; ‘4’ indicates a backbone
CO · NH hydrogen bond; ‘5’ indicates the
presence of both these backbone hydrogen bonds. ‘6’,
‘7’ and ‘8’ indicate interactions similar
to ‘3’, ‘4’ and ‘5’ except
they also indicate the presence of a heavy atom interaction.
The PSF database is based on the PDB version of August 2008.
It is expected, that this database will be manually updated
every 6 months, until an automatic update is programmed. The
database was prepared using Perl scripts and the C++ DSSP script
(
11).
|
SEARCH ENGINE ALGORITHM
|
|---|
To extract segments obeying the query parameters from the aforementioned
database, a search engine was designed using Perl. The engine
parses over the entire database and searches for matches of
amino acid sequence, secondary structure and contacts in a procedural
manner. First, the program attempts to match the amino acid
sequence. If successful, the program proceeds and attempts to
find a secondary structure match, else the next database entry
is read. If also the secondary structure is matched then the
program continues to find contact matches, else the next database
entry is read. This cycle is repeated until all database entries
are read. All matches of sequence, secondary structure and contact
are stored in three separate lists. These three lists are then
compared for common sequence position of the matched segments.
If a sequence position is identical in all three lists, then
the PDB ID and sequence position of the match is stored in a
final hit list. This hit list information is then forwarded
to the output manager which in turn displays it textually and
graphically.
|
GRAPHIC USER INTERFACE
|
|---|
To facilitate public use of the program, a GUI was designed
using Javascript. In this GUI, the user is prompted to select
a query type, from the following: segment size, amino acid sequence,
secondary structure sequence, secondary structure element and
contacts (
Figure 2). Querying for segment size prompts the user
to select a length between 1 and 30 residues. Querying for an
amino acid or a secondary structure sequence prompts the user
to enter the sequence in the one letter code. Querying for a
secondary structure element prompts the user to choose among
predefined ones such as
-helix,
-hairpin, β-strand, β-hairpin,
random coil, turn,
and 3
10-helices.
View larger version (34K):
[in this window]
[in a new window]
[Download PowerPoint slide]
|
Figure 2. Representative query example. Shown is the query form for a segment of 12 residues that adopts a β-hairpin conformation with the sequence CxCxxxxxCxC (‘x’ represents any amino acid), two disulfide bonds (in yellow), four hydrogen bonds (in blue) and one heavy atom interaction (in red) as drawn.
|
|
Selecting a query for contacts launches a simplistic molecular
display of the queried segment in which contacts are added by
mouse clicking on interacting atom pairs (
Figure 2). Up to four
types of contacts may be added in this manner, namely heavy
atom contacts by clicking on the R groups of interacting residues
(
Figure 2), C
–C
interactions by clicking on C
pairs, backbone
hydrogen bonds by clicking on HN and CO atoms, and disulfide
bridges by clicking on sulfur atom pairs. Note that because
residue contacts are defined as any pair of heavy atoms in the
two residues being closer than 6 Å, the atom marker ‘R’
is essentially any heavy atom in the residue. Upon clicking
on the atom pairs a burgundy, red, blue or yellow line connecting
the interacting atoms will appear representing C
, heavy atoms,
hydrogen bonds and disulfide contacts, respectively.
For each query type an estimation of the number of matches is available for preview by clicking on the ‘evaluate subquery’ button. The ‘evaluate sub-query’ is not a prerequisite for job submission, but rather a means to estimate the number of segment that match the given constraints. An estimate of the total number of hits is obtainable by clicking on the ‘evaluate’ button. To include more or less query types, click the ‘+’ and ‘–’ button, respectively. For expediency, example queries have been prepared and are available by clicking the ‘example’ buttons. A tutorial and help messages are available online.
|
DATA RETRIEVAL
|
|---|
After entering all the information, the user can run the query
by clicking the ‘submit’ button. The query parameters
are then passed on to the search engine using CGI and the run
is initiated. Upon run completion, query matches are displayed
in text and figures (
Figure 3). Text output includes primary
and secondary sequence, as well as the relevant contact map.
Graphic output includes a Rasmol (
12) generated image or a Jmol
interactive molecular viewer applet allowing easy viewing of
the matched segments. To allow handling of the voluminous results,
an adjustable paging system is enabled. The GUI does not require
any prior knowledge of scripting languages and permits public
and general use of the engine. An alternative to the GUI is
a nongraphic text interface, which is available by clicking
on the ‘text only’ button in the output format.
This featured interface offers an easy solution for browsers
not supporting Javascript. Finally, the database content and
its contact map may be retrieved by entering the PDB ID in the
query type option ‘retrieve PDB entry’.
View larger version (38K):
[in this window]
[in a new window]
[Download PowerPoint slide]
|
Figure 3. Representative output example. Shown is the output of the query example (Figure 2) for the amino acid sequence CxCxxxxxCxC in a β-hairpin conformation. The additional query parameters (two disulfide bridges, four hydrogen bonds and one heavy atom contact) are summarized in the header of the output. Only the three first hits of a total of 68 are shown. For each hit, the output display includes the PDB ID, the sequence position and chain, the amino acid sequence, the secondary structure, the relevant contact map and a Jmol interactive molecular viewer applet for viewing the segment structure.
|
|
|
SERVER CONFIGURATION
|
|---|
The search engine described above is publicly available at
http://ari.stanford.edu/psf as a community service. The engine runs on a small server consisting
of a Linux operated desktop computer with a 1.8 GHz Intel Pentium
processor, with 2 GB of RAM. The web interface is powered by
Apache HTTP server version 2.2. Typical search durations are
<20 s.
|
PREDEFINED SECONDARY STRUCTURE ELEMENTS
|
|---|
As mentioned earlier, it is possible to select predefined secondary
structure elements, such as β-hairpin, β-strand,
-hairpin,
,
and 3
10-helices. These predefined elements are simply defined
as multiple repeats of the basic secondary structure unit, with
the exception of
- and β-hairpins. The length of these
elements is set by choosing an appropriate ‘segment size’
between 1 and 30 residues. The program defines β-hairpins
as a segment containing a secondary structure of the form EXE
in which E is a β-strand and X is a gap, a bend, or a turn,
and displaying at least one hydrogen bond at its base. For segments
of even length,
n 
4, the β-hairpin secondary structure
is defined by the following palindromic regular expression E{1}(E|T|S|\s){(
n – 4)/2}(T|S|\s){2}(E|T|S|\s){(
n – 4)/2}E{1}. For
segments with odd number length,
n 3, the β-hairpin secondary
structure is defined by the following palindromic regular expression
E{1}(T|S|\s){(
n – 3)/2}(T|S|\s){1}(T|S|\s){(
n –
3)/2}E{1}. To ensure β-hairpin closure and adequate hydrogen
bonding, the first and last residues of the hairpin are connected
by at least one hydrogen bond. This hydrogen bond may be erased
when defining contacts and other hydrogen bonds of choice may
be added instead. Notably, this hydrogen bond may be moved by
one index, thus inverting the β-hairpin by clicking on
the button labeled ‘reverse hydrogen-bond network’.
Correspondingly, -hairpins are defined by the program as a segment containing a secondary structure of the form HXH in which H is an -helix and X represents a gap, a bend or a turn, with at least one contact between the helices. For segments of even length, n 4, the secondary structure is defined as H{1}(H|T|S|\s){(n – 4)/2}(T|S|\s){2}(H|T|S|\s){(n – 4)/2}H{1}, and for segments of odd number length, n 3, the secondary structure is defined as H{1}(H|T|S|\s){(n – 3)/2}(T|S|\s){1}(H|T|S|\s){(n – 3)/2}H{1}. To ensure proper alignment, the first and last residues of the -hairpin have at least one contact. This contact may be erased when defining contacts. Among the strengths of the search engine is the ability to identify - and β-hairpins in a rapid and effective manner.
|
CONCLUSIONS
|
|---|
Over the past decade, we have witnessed the development of methods
for fold classification such as SCOP (
13) and CATH (
14). More
recently, however, attention has shifted to structural similarities
at an atomic level rather than of the domain fold, such as the
conformation of protein segments. Whereas the overall fold is
indisputably significant as a framework upon which protein function
lays, the actual protein function is usually carried out by
a relatively small amount of residues or a protein segment.
The search engine presented herein allows the easy and quick
identification of such conformational segments. It is expected
to be beneficial to the scientific community for comparative
structural analysis, for the analysis of protein segments, and
for the prediction of structure and function of uncharacterized
proteins. The search engine is particularly valuable for finding
homologs of NMR structures, which are based on a large number
of contacts. In the future, we anticipate a publicly available
PSF program for download and use on the client side. We also
intend to improve the search engine by allowing search for RNA
motifs as well as larger protein segments, thus enabling a more
comprehensive survey.
|
FUNDING
|
|---|
National Institutes of Health (GM63817 and EY016525
[GenBank]
). Funding
for open access charge: National Institutes of Health (GM63817).
Conflict of interest statement. None declared.
|
ACKNOWLEDGEMENT
|
|---|
We thank Prof. Jacob Anglister and Dr Osnat Rosen for testing
PSF.
|
REFERENCES
|
|---|
- Levitt M. Growth of novel protein structural data. Proc. Natl Acad. Sci. USA (2007) 104:3183–3188.[Abstract/Free Full Text]
- Levitt M. Accurate modeling of protein conformation by automatic segment matching. J. Mol. Biol. (1992) 226:507–533.[CrossRef][Web of Science][Medline]
- Berman HM, Battistuz T, Bhat TN, Bluhm WF, Bourne PE, Burkhardt K, Feng Z, Gilliland GL, Iype L, Jain S, et al. The Protein Data Bank. Acta Crystallogr. D Biol. Crystallogr. (2002) 58:899–907.[CrossRef][Medline]
- Golovin A, Henrick K. MSDmotif: exploring protein sites and motifs. BMC Bioinformatics (2008) 9:312.[CrossRef][Medline]
- Kleywegt GJ. Recognition of spatial motifs in protein structures. J. Mol. Biol. (1999) 285:1887–1897.[CrossRef][Web of Science][Medline]
- Ananthalakshmi P, Kumar Ch K, Jeyasimhan M, Sumathi K, Sekar K. Fragment Finder: a web-based software to identify similar three-dimensional structural motif. Nucleic Acids Res. (2005) 33:W85–W88.[Abstract/Free Full Text]
- Taubig H, Buchner A, Griebsch J. PAST: fast structure-based searching in the PDB. Nucleic Acids Res. (2006) 34:W20–W23.[Abstract/Free Full Text]
- Kato H, Takahashi Y. SS3D-P2: a three dimensional substructure search program for protein motifs based on secondary structure elements. Comput. Appl. Biosci. (1997) 13:593–600.[Abstract/Free Full Text]
- Akutsu T, Onizuka K, Ishikawa M. Rapid protein fragment search using hash functions based on the Fourier transform. Comput. Appl. Biosci. (1997) 13:357–364.[Abstract/Free Full Text]
- Ausiello G, Zanzoni A, Peluso D, Via A, Helmer-Citterich M. pdbFun: mass selection and fast comparison of annotated PDB residues. Nucleic Acids Res. (2005) 33:W133–W137.[Abstract/Free Full Text]
- Kabsch W, Sander C. Dictionary of protein secondary structure: pattern recognition of hydrogen-bonded and geometrical features. Biopolymers (1983) 22:2577–2637.[CrossRef][Web of Science][Medline]
- Sayle RA, Milner-White EJ. RASMOL: biomolecular graphics for all. Trends Biochem. Sci. (1995) 20:374.[CrossRef][Web of Science][Medline]
- Murzin AG, Brenner SE, Hubbard T, Chothia C. SCOP: a structural classification of proteins database for the investigation of sequences and structures. J. Mol. Biol. (1995) 247:536–540.[CrossRef][Web of Science][Medline]
- Orengo CA, Michie AD, Jones S, Jones DT, Swindells MB, Thornton JM. CATH–a hierarchic classification of protein domain structures. Structure (1997) 5:1093–1108.[Medline]
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
|
|
 
Y. Yamazaki, R. Akashi, Y. Banno, T. Endo, H. Ezura, K. Fukami-Kobayashi, K. Inaba, T. Isa, K. Kamei, F. Kasai, et al.
NBRP databases: databases of biological resources in Japan
Nucleic Acids Res.,
November 24, 2009;
(2009)
gkp996v1.
[Abstract]
[Full Text]
[PDF]
|
|
|
|
|
|
|
D. S. Berkholz, P. B. Krenesky, J. R. Davidson, and P. A. Karplus
Protein Geometry Database: a flexible engine to explore backbone conformations and their relationships to covalent geometry
Nucleic Acids Res.,
November 11, 2009;
(2009)
gkp1013v1.
[Abstract]
[Full Text]
[PDF]
|
|
|
TOP
ABSTRACT
INTRODUCTION