Vol. 30 no.

6 2014, pages 889–890

BIOINFORMATICS APPLICATIONS NOTE doi:10.1093/bioinformatics/btt645

Structural bioinformatics Advance Access publication November 7, 2013

PREDDIMER: a web server for prediction of transmembrane

helical dimers
Anton A. Polyansky1,2,*, Anton O. Chugunov1, Pavel E. Volynsky1, Nikolay A. Krylov1,
Dmitry E. Nolde1 and Roman G. Efremov1,3

Downloaded from [Link] by University of Valencia user on 10 June 2025

1
M.M. Shemyakin and Yu.A. Ovchinnikov Institute of Bioorganic Chemistry, Russian Academy of Sciences, Moscow
117997, Russia, 2Max F. Perutz Laboratories, Department of Structural and Computational Biology, University of Vienna,
Campus Vienna Biocenter 5, Vienna AT-1030, Austria and 3Moscow Institute of Physics and Technology (State
University), Dolgoprudny, Moscow Region 141700, Russia
Associate Editor: Anna Tramontano

ABSTRACT techniques of TM dimer structures determination (Bocharov

Summary: Here we present PREDDIMER, a web tool for prediction of et al., 2010), the existing modeling approaches vary a lot in
dimer structure of transmembrane (TM) helices. PREDDIMER allows (i) their complexity, prediction power and computational efficiency:
reconstruction of a number of dimer structures for given sequence(s) starting from simple sequence motif-based algorithms to labori-
of TM protein fragments, (ii) ranking and filtering of predicted struc- ous force-field techniques like Monte Carlo conformational
tures according to respective values of a scoring function, (iii) visual- search in implicit membranes, coarse-grained or atomic molecu-
ization of predicted 3D dimer structures and (iv) visualization of surface lar dynamics simulations in lipid bilayers or different combin-
hydrophobicity of TM helices and their contacting (interface) regions ations of them (Polyansky et al., 2011). Here, we present
represented as 2D maps. PREDDIMER web server for prediction of ensembles of
Results: We implemented online the original PREDDIMER algorithm
possible spatial structures of TM helical dimers based just on
respective TM sequence(s). Our web server uses original
and benchmarked the server on 11 TM sequences, whose 3D dimer
PREDDIMER algorithm described elsewhere (Polyansky
conformations were obtained previously by nuclear magnetic reson-
et al., 2012). To our knowledge up-to-date, this is the only online
ance spectroscopy. In the most of tested cases backbone root-
tool, which allows fast and efficient prediction of TM dimers
mean-square deviations of closest predicted conformations from the
structure as well as visualization of predicted results both in
experimental reference are below 3 Å. A randomization test displays
3D and 2D (2D maps of surface hydrophobicity of interacting
good anticorrelation ( 0.82) between values of the scoring function and
helices with outlined contacting interfaces). Although the server
statistical significance of the prediction ‘by chance’. Going beyond a does not address a question whether two given TM fragments
single dimer conformation, our web tool predicts an ensemble of pos- dimerize or not, which is still far from being fully reachable for
sible conformations, which may be useful for explanation of a function- solely in silico techniques, it provides their putative dimer con-
ing of bitopic membrane proteins, e.g. receptor tyrosine kinases. formations along with estimation of their reliability.
Availability and implementation: PREDDIMER can be accessed for
free on the web at [Link]
Contact: newant@[Link] 2 WORKFLOW AND IMPLEMENTATION
Supplementary information: Supplementary data are available at The PREDDIMER server has two modes: (i) prediction of TM-dimer
Bioinformatics online. structures from sequences and (ii) analysis of hydrophobic properties and
contacting regions for an existing dimer structure (separate Protein Data
Received on September 2, 2013; revised on October 21, 2013;
Bank (PDB) files for each of interacting helices have to be uploaded by
accepted on November 3, 2013 user). In the first case, user should input one or two TM sequences (single
for homo- and two for heterodimers, respectively) with a length 20–35
amino acid residues, optionally specify their name(s) and indicate orienta-
1 INTRODUCTION tion of helices in a dimer (parallel/antiparallel) and pH, which determines
Transmembrane (TM) domains of bitopic proteins comprise just ionization state of amino acid sidechains. On submitting a prediction task,
a single helix. For particular class of such proteins with high the server sends an e-mail notification with unique link to a result page,
biological importance—receptor tyrosine kinases (RTKs)—it which will be created on calculations completion. On average, the predic-
has been shown that TM fragments play a crucial role in dimer- tion takes 20 min depending on the server workload. The server is im-
ization and activation of these receptors by controlling orienta- plemented using Python and PHP programming languages, and also Zend
tion of their intracellular kinase domains (Cymer and Schneider, and Bootstrap frameworks. The algorithm details are available in online
manual ([Link] Parameters of predicted
2010). The knowledge of spatial structures of dimers of TM frag-
dimer conformations such as rank of a structure, FSCOR value, helix–helix
ments of bitopic proteins is important for understanding the
crossing angle () and rotation angles around helical axes ( 1, 2) are
functioning of these molecules upon their oligomerization in presented in an interactive table, where selection of a certain row results
the cell membrane. Being reasonable alternative to experimental in output of respective 2D maps for both helices and interactive 3D con-
formation of the dimer. Predicted PDB structures, 2D maps (PNG and
*To whom correspondence should be addressed. PDF) and resulting table (text format) are available for download as

ß The Author 2013. Published by Oxford University Press. All rights reserved. For Permissions, please e-mail: [Link]@[Link] 889
[Link] et al.

separate files or combined ZIP archive. Similar result page can be created Because the algorithm operates with ideal helices, it outputs
for an existing dimer structure using analyzing mode (mode 2) of the rather ‘coarse-grained’ initial dimer configurations, whereas the
server. membrane environment is not accounted explicitly. At the same
time, the predicted conformations represent a reliable starting
point for further molecular dynamics optimization in explicit
3 PERFORMANCE membranes, which usually increases their similarity to the experi-
mental structures (Polyansky et al., 2012). Noteworthy, the
We calculated possible conformations for 10 homodimers and 1
algorithm predicts a number of possible dimer configurations,
heterodimer, whose structures were obtained previously by
which is not a case of NMR techniques usually giving a set of
nuclear magnetic resonance (NMR) spectroscopy (Fig. 1A).

Downloaded from [Link] by University of Valencia user on 10 June 2025

similar models optimally suited for the particular membrane
In most cases, NMR-like conformations are among three top-
mimic (micelles, bicelles). The calculated dimer conformations
scoring models for a given dimer (total number of models is
can correspond to different intermediate states of TM domains,
shown in parentheses) and display backbone root-mean-square
thus shedding light on the mechanism of protein functioning. We
deviations (RMSD) from the references below 3 Å. The only
illustrate this on the example of RTK ErbB2 (Fig. 1B and C).
exceptions are ErbB3 homodimer, whose published structure dis-
The first predicted right-handed model (Fig. 1B, left) corres-
plays anomalous packing compared with other RTKs (Li et al.,
ponds well to the experimental structure (2JWA, RMSD of
2012), and FGFR3, whose NMR structure is characterized by
1.43Å), whereas the second one (Fig. 1B, right) is similar to
the presence of distorted conformations of TM helices (Volynsky
alternative left-handed state suggested by modeling for this
et al., 2013). For the most well-known TM dimer structure of
dimer (Fleishman et al., 2002). Switching between possible inter-
glycophorin A (GpA, PDB: 1AFO), the server gives a top-
action interfaces (Fig. 1C) results in different helices orientation
ranked model with RMSD of 1.56 Å from the reference. We
with respect to intracellular kinase domains and may contribute
performed also randomization test, where probability to form
to activation of ErbB2 (Fleishman et al., 2002) and other RTKs
a dimer between a given TM fragment and random ones were
(Endres et al., 2013; Volynsky et al., 2013). In this context,
estimated over 1000 shuffles of interacting sequence. P-values
PREDDIMER web tool goes beyond common concept of just
obtained for the studied dimer models display good anticorrel-
a single structure of TM helical dimer usually available experi-
ation ( 0.82) with the respective FSCOR values (Supplementary
mentally in specific conditions of NMR setup.
Figure S1), suggesting the scoring function as reasonable and fast
estimate of the reliability of predicted conformations.

ACKNOWLEDGEMENTS
The authors thank Sergey Goncharuk for the help with server
setup.
Funding: Ministry of Education and Science of the Russian
Federation (14.514.11.4069), the Russian Foundation for Basic
Research and the RAS Programmes ‘Basic fundamental research
for nanotechnologies and nanomaterials’ and ‘Molecular and
Cellular Biology’.
Conflict of Interest: none declared.

REFERENCES
Bocharov,E.V. et al. (2010) Structure elucidation of dimeric transmembrane do-
mains of bitopic proteins. Cell Adh. Migr., 4, 284–298.
Cymer,F. and Schneider,D. (2010) Transmembrane helix-helix interactions involved
in ErbB receptor signaling. Cell Adh. Migr., 4, 299–312.
Efremov,R.G. et al. (2007) Molecular lipophilicity in protein modeling and drug
design. Curr. Med. Chem., 14, 393–415.
Endres,N.F. et al. (2013) Conformational coupling across the plasma membrane in
activation of the EGF receptor. Cell, 152, 543–556.
Fleishman,S.J. et al. (2002) A putative molecular-activation switch in the transmem-
brane domain of erbB2. Proc. Natl Acad. Sci. USA, 99, 15937–15940.
Fig. 1. Prediction results for TM helical dimers obtained using Li,E. et al. (2012) Transmembrane helix dimerization: beyond the search for se-
PREDDIMER server. (A) Parameters of dimer conformations compared quence motifs. Biochim. Biophys. Acta, 1818, 183–193.
with the experimental references (see Supplementary Material for details). Polyansky,A.A. et al. (2011) Structural, dynamic, and functional aspects of helix
association in membranes: a computational view. Adv. Protein Chem. Struct.
(B) Spatial structures of NMR-like (left) and alternative (right) models for
Biol., 83, 129–161.
ErbB2. Peptides are shown in cartoon-and-sticks representation. (C) 2D
Polyansky,A.A. et al. (2012) Multistate organization of transmembrane helical pro-
maps of the molecular hydrophobicity potential (MHP) (Efremov et al., tein dimers governed by the host membrane. J. Am. Chem. Soc., 134,
2007) on the peptide surfaces in two ErbB2 models generated by the server. 14390–14400.
Dimerization interfaces are outlined. Axis values correspond to the rotation Volynsky,P.E. et al. (2013) Role of dimerization efficiency of transmembrane
angle around the helical axis ( ) and the distance along the latter (Z), domains in activation of fibroblast growth factor receptor 3. J. Am. Chem.
respectively. 2D maps are colored according to MHP arbitrary units Soc., 135, 8105–8108.

890