Rapid membrane protein topology prediction.
ABSTRACT State-of-the-art methods for topology of α-helical membrane proteins are based on the use of time-consuming multiple sequence alignments obtained from PSI-BLAST or other sources. Here, we examine if it is possible to use the consensus of topology prediction methods that are based on single sequences to obtain a similar accuracy as the more accurate multiple sequence-based methods. Here, we show that TOPCONS-single performs better than any of the other topology prediction methods tested here, but ~6% worse than the best method that is utilizing multiple sequence alignments. AVAILABILITY AND IMPLEMENTATION: TOPCONS-single is available as a web server from http://single.topcons.net/ and is also included for local installation from the web site. In addition, consensus-based topology predictions for the entire international protein index (IPI) is available from the web server and will be updated at regular intervals.
- [Show abstract] [Hide abstract]
ABSTRACT: Most integral membrane proteins, both in prokaryotic and eukaryotic cells, are co-translationally inserted into the membrane via Sec-type translocons: the SecYEG complex in prokaryotes and the Sec61 complex in eukaryotes. The contributions of individual amino acids to the overall free energy of membrane insertion of single transmembrane -helices have been measured for Sec61-mediated insertion into the endoplasmic reticular (ER) membrane (Nature 450:1026-1030), but have not been systematically determined for SecYEG-mediated insertion into the bacterial inner membrane. We now report such measurements, carried out in Escherichia coli. Overall, there is a good correlation between the results found for the mammalian ER and the E. coli inner membrane, but the hydrophobicity threshold for SecYEG-mediated insertion is distinctly lower than for Sec61-mediated insertion.Journal of Molecular Biology 05/2013; · 3.91 Impact Factor
- [Show abstract] [Hide abstract]
ABSTRACT: Topology analysis of membrane proteins can be obtained by enzymatic shaving in combination with mass-spectrometry identification of peptides. Ideally, such analysis could provide quite detailed information about the membrane spanning regions. Here, we examine the ability of some shaving enzymes to provide large-scale analysis of membrane proteome topologies. To compare different shaving enzymes, we first analyzed the detected peptides from two over-expressed proteins. Secondly, we analyzed the peptides from non over-expressed Escherichia coli membrane proteins with known structure to evaluate the shaving methods. Finally, the identified peptides were used to test the accuracy of a number of topology predictors. At the end we suggest that the usage of thermolysin, an enzyme working at the natural pH of the cell for membrane shaving is superior because: (i) we detect a similar number of peptides and proteins using thermolysin and trypsin. (ii) Thermolysin shaving can be run at a natural pH and (iii) the incubation time is quite short. (iv) Fewer detected peptides from thermolysin shaving originate from the transmembrane regions. Using thermolysin shaving we can also provide a clear separation between the best and the less accurate topology predictors, indicating that using data from shaving can provide valuable information when developing new topology predictors.Proteomics 03/2013; · 4.43 Impact Factor
- [Show abstract] [Hide abstract]
ABSTRACT: During the last two decades a large number of computational methods have been developed for predicting transmembrane protein topology. Current predictors rely on topogenic signals in the protein sequence, such as the distribution of positively charged residues in extra-membrane loops and the existence of N-terminal signals. However, phosphorylation and glycosylation are post-translational modifications (PTMs) that occur in a compartment-specific manner and therefore the presence of a phosphorylation or glycosylation site in a transmembrane protein provides topological information. We examine the combination of phosphorylation and glycosylation site prediction with transmembrane protein topology prediction. We report the development of a Hidden Markov Model based method, capable of predicting the topology of transmembrane proteins and the existence of kinase specific phosphorylation and N/O-linked glycosylation sites along the protein sequence. Our method integrates a novel feature in transmembrane protein topology prediction, which results in improved performance for topology prediction and reliable prediction of phosphorylation and glycosylation sites. The method is freely available at http://aias.biol.uoa.gr/HMMpTM.Biochimica et Biophysica Acta 11/2013; · 4.66 Impact Factor
BIOINFORMATICS APPLICATIONS NOTE
Vol. 27 no. 9 2011, pages 1322–1323
Rapid membrane protein topology prediction
Aron Hennerdal and Arne Elofsson∗
Department of Biochemistry and Biophysics, Stockholm Bioinformatics Center, Center for Biomembrane Research,
Swedish e-science Research Center, Stockholm University, 106 91 Stockholm, Sweden
Associate Editor: Burkhard Rost
Summary: State-of-the-art methods for topology of α-helical
membrane proteins are based on the use of time-consuming multiple
sequence alignments obtained from PSI-BLAST or other sources.
Here, we examine if it is possible to use the consensus of topology
prediction methods that are based on single sequences to obtain
a similar accuracy as the more accurate multiple sequence-based
methods. Here, we show that TOPCONS-single performs better
than any of the other topology prediction methods tested here, but
∼6% worse than the best method that is utilizing multiple sequence
Availability and Implementation: TOPCONS-single is available as
a web server from http://single.topcons.net/ and is also included for
local installation from the web site. In addition, consensus-based
topology predictions for the entire international protein index (IPI) is
available from the web server and will be updated at regular intervals.
Supplementary information: Supplementary data are avaliable at
Received on December 2, 2010; revised on February 16, 2011;
accepted on February 28, 2011
Today only 268 unique α-helical membrane protein structures are
known according to the Orientation of Proteins in Membranes
database (OPM, http://opm.phar.umich.edu/). The ‘topology’ of
such proteins has proven to be a convenient concept. In essence,
the topology specifies the number of transmembrane α-helices of
the protein together with the location of the N-terminal end of the
chain, i.e. whether it is in the cytosol (‘in’) or in the endoplamsatic
reticulum (ER) lumen or extramembrane space (‘out’).
The TOPCONS algorithm (Bernsel et al., 2009) computes
Markov Model (HMM) and input from several topology predictors.
and is based on five state-of-the-art topology prediction methods
and typically takes a couple of minutes to run. The bulk of that
time is spent running a PSI-BLAST (Altschul et al., 1997) search
against a sequence database to obtain evolutionary information that
is then used by the underlying predictors. This approach is quite
accurate, but woefully inappropriate when running predictions for
many sequences, e.g. in studies of whole genomes.
∗To whom correspondence should be addressed.
Table 1. The accuracy of different predictors on different datasets
Homology reduced to 30% sequence identity. The numbers in parenthesis denote the
number of protein sequences in the set. ‘Time’ is the time it takes to process the set of
101 protein sequences.
Here, we have benchmarked the TOPCONS algorithm (Bernsel et al., 2009)
information, i.e. do not require BLAST to be run. Six individual methods
were tested: SCAMPI-single (Bernsel et al., 2008) S-TMHMM (Viklund
and Elofsson, 2004), HMMTOP (Tusnády and Simon, 2001), TopPred (von
Heijne, 1992; Claros and Heijne, 1994), MEMSAT-1.0 (Jones et al., 1994)
and PHOBIUS (Käll et al., 2004).
The methods were benchmarked using a modified version of the dataset
used in SCAMPI (Bernsel et al., 2008). The original set consisted of
two subsets stemming from the high-resolution structures (123 sequences)
and from structures of lower resolution (146 sequences). This set was
homology-reduced to 30% sequence identity using the method proposed
and implemented by Holm and Sander (1998). The reduced set contain 101
sequences and was further divided into multi-spanning (79 sequences) and
single-spanning (22 sequences) proteins resulting in three sets labeled ‘all’,
‘multi’ and ‘single’, respectively.
All possible combinations of three or more topology predictors were used
as input to the TOPCONS algorithm and the results were evaluated. The
best combination—the one scoring the highest accuracy over the dataset—is
listed in Table 1. Accuracy is the proportion of correct predictions, and
correct topology predictions are defined as by Krogh et al. (2001). All
definitions and the full list of all method combinations are available in
the Supplementary Material. To enable comparison, the performance of the
original TOPCONS server based on homology information is listed, as well
as the individual performance for the six single sequence methods.
The execution time for each run of TOPCONS-single was also measured.
For the benchmark dataset, the time required from start to finish varied
between 50s (when using three methods) to 100s (when using all six
needed 4500s and the fastest individual method, SCAMPI-single, finished
in ∼2s. Running the complete human genome (∼21000 sequences) through
the TOPCONS-single pipeline took ∼60min.
DEVELOPMENT OF TOPCONS-SINGLE
© The Author(s) 2011. Published by Oxford University Press.
This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/
by-nc/2.5), which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited.
Rapid membrane protein topology prediction
Fig. 1. Coverage versus correct topology predictions for TOPCONS-single
and each of the individual methods. The proteins in the test set (‘all’) are
ordered according to the decreasing reliability score, and the percentage of
correct predictions are calculated every 10% of coverage.
We have implemented a reliability score for TOPCONS-single as
individual method, as previously described (Melén et al., 2003). Definitions
and descriptions of all reliability scores are listed in the Supplementary
Material. We investigated the reliability scores by ranking the predictions by
descending reliability score and plotting the fraction of correct predictions
against the coverage in the ‘All’ benchmark dataset (Fig. 1).
We have constructed a consensus predictor for α-helical membrane
proteins using the HMM-based TOPCONS algorithm with several
fast single sequence-based prediction methods as input. After
starting out with six predictors and benchmarking all possible
combinations and subsets of them, we found that a combination of
the best results. TOPCONS-single consistently outperforms each
of its underlying single sequence predictors when they are used
on their own, which confirms the notion of consensus prediction
adding value. It does not use searches for homologous proteins and
thus performs worse, but runs much faster than a corresponding
approach using evolutionary information.
TOPCONS-single performs especially well on single-spanning
membrane proteins in our dataset (Table 1) mainly by not over-
predicting the number of transmembrane helices in the same extent
as the single sequence methods (Supplementary Material).
A possible caveat to our approach is the use of benchmark
sets where at least subsets have been previously used to train
the underlying single sequence methods. We judge this to be less
influential since the authors of said prediction methods have taken
steps to avoid overtraining on their respective sets.
The best-performing version of TOPCONS-single, using four
individual methods (Table 1), is available as an easy-to-use web-
based prediction server at http://single.topcons.net/. It uses the
globular protein filter of SCAMPI to weed out non-membrane
proteins and then proceeds to run the rest of the predictors–
HMMTOP, MEMSAT-1.0 and S-TMHMM—on the remaining set.
The output consists of text files with well-defined formats for easy
We would like to thank Dr. Håkan Viklund for writing the modhmm
code used in TOPCONS.
Funding: This work was supported by grants from the Swedish
Research Council (VR-NT 2009-5072, VR-M 2007-3065), SSF
(the Foundation for Strategic Research), the EU 6’th Framework
Program by support to the EMBRACE project (contract No:
LSHG-CT-2004-512092) and the EU 7’th Framework Program by
Conflict of Interest: none declared.
Altschul,S.F. et al. (1997) Gapped blast and psi-blast: a new generation of protein
database search programs. Nucleic Acids Res., 25, 3389–3402.
Bernsel,A. et al. (2008) Prediction of membrane-protein topology from first principles.
Proc. Natl Acad. Sci. USA, 105, 7177–7181.
Bernsel,A. et al. (2009) Topcons: consensus prediction of membrane protein topology.
Nucleic Acids Res., 37, W465–W468.
Claros,M.G. and von Heijne,G. (1994) TopPred II: an improved software for membrane
protein structure predictions. Comput. Appl. Biosci., 10, 685–686.
sequence collections. Bioinformatics, 14, 423–429.
Jones,D.T. et al. (1994) A model recognition approach to the prediction of all-helical
membrane protein structure and topology. Biochemistry, 33, 3038–3049.
Käll,L. et al. (2004)Acombined transmembrane topology and signal peptide prediction
method. J. Mol. Biol., 338, 1027–1036.
model: application to complete genomes. J. Mol. Biol., 305, 567–580.
Melén,K. et al. (2003) Reliability measures for membrane protein topology prediction
algorithms. J. Mol. Biol., 327, 735–744.
Tusnády,G.E. and Simon,I. (2001) The hmmtop transmembrane topology prediction
server. Bioinformatics, 17, 849–850.
Viklund,H. and Elofsson,A. (2004) Best alpha-helical transmembrane protein topology
Protein Science, 13, 1908–1917.
von Heijne,G. (1992) Membrane protein structure prediction. hydrophobicity analysis
and the positive-inside rule. J. Mol. Biol., 225, 487–494.