Bioinformatics of membrane proteins

1
Bioinformatics of membrane proteins
Gunnar von Heijne Department of Biochemistry and
Biophysics Stockholm Bioinformatics
Center Stockholm University
2
I. The physical view
3
A simulated lipid bilayer(Jakobsson, TiBS 22339)
Physical properties
4
Physcial properties of the lipid bilayer (White
et al., JBC 27632395)
interactrions
5
Molecular interactions in membrane
proteins(White et al., JBC 27632395)
2 structures
6
Only two basic structures(Quart.Rev.Biophys.
32285)
ß-barrel
Lipid/prot interactions
7
Lipid-protein interactions(Mitsuoka et al., JMB
286861)
Exposed/buried
8
Lipid-exposed vs. buried residues(Prot. Sci.
6808)
conservation
9
Lipid-exposed residues are less conserved than
buried ones
Membrane assembly
10
II. Assembly in vivo
11
Protein sorting in a eukaryotic cell
SRP pwy
12
The SRP/Sec61 pathway
Ffh
13
The Ffh M-domain/4.5 S complex(Batey et al.,
Science 2871232)
ribosome
14
The nascent chain tunnel(Nissen et al., Science
289920)
Rib-Sec61
15
The ribosome-Sec61p complex(Beckmann et al.,
Science 2782123)
16
The translocation channel(Beckmann et al.,
Science 2782123)
movie
17
The basic model(graphics by Bill Skach)
prediction
18
III. Prediction - basics
19
What we want
- know all membrane proteins - know their
topology - know their 3D structure - know their
function and more...
TM lengths
20
TM helix lengths are typically 20-30
residues(Bowie, JMB 272780)
Trp, Tyr
21
Trp Tyr are enriched in the region near the
lipid headgroups(Prot.Sci. 6808 72026)
Loop lengths
22
Loops tend to be short(Tusnady Simon, JMB
283489)
PI rule
23
The positive inside rule(EMBO J. 53021 EJB
174671, 2051207 FEBS Lett. 28241)
Bacterial IM in 16 KR out 4 KR Eukaryotic
PM in 17 KR out 7 KR Thylakoid membrane in
13 KR out 5 KR Mitochondrial IM In 10 KR
out 3 KR
out
in
prediction
24
IV. Topology prediction
25
Topology prediction - a classical problem in
bioinformatics
4 characteristics
26
Four important characteristics
short loops
20 hydrophobic residues
Positive inside rule
predictors
27
Popular topology predictors
TMHMM (HMM) HMMTOP (HMM) TopPred (h-plot
PI-rule) MEMSAT (dynamic programming) TMAP
(h-plot, mult. alignment) PHD (NN, mult.
alignment)
toppred
28
TopPred(JMB 225487)
- construct all possible topologies - rank based
on D
E. coli LacY
http//bioweb.pasteur.fr/ seqanal/interfaces/ topp
red.html
TMHMM
29
TMHMM(Sonnhammer et al., ISMB 6175, Krogh et
al., JMB 305567)
A hidden Markov model-based method
www.cbs.dtu.dk
h l models
30
Helix loop models in TMHMM
HMMTOP
31
TMHMM performance(Krogh et al., JMB 305567)
Discrimination globular/membrane sens spec gt
98 Correct topology 65-70 Single TM
identification sensitivity 96 specificity
98 Training set 160 membrane proteins 650
globular proteins
of TM proteins
32
How many membrane proteins are there?
- identify all TMHMM hits with TM 1 - remove
secretory proteins from the 1TM class using
SignalP-HMM
results
33
20-25 of all ORFs encode membrane proteins
C. elegans 30 D. melanogaster 20 S.
cerevisiae 21 A. thaliana 23 B.
subtilis 24 E. coli 21 M. genitalium 20
T. maritima 24 A. fulgidus 20 P.
horikoshii 26
consensus
34
Consensus predictions indicate
reliability(FEBS Lett. 486267)
60 E. coli proteins
5 prediction methods used 46 of 764 predicted E.
coli IM proteins are in the 5/0 or 4/1 classes
fraction correct/coverage
majority level
Partial consensus
35
Partial consensus topologies(Prot. Sci. 112974)
- 89 of all predicted partial topologies
correct (5/0 class) - 72 of all predicted E.
coli IM proteins covered (5/0 class)
TMHMM reliability
36
TMHMM reliability scores(Melén et al., JMB in
press)
TMHMM output
1. Mean probability pmean 2. Minimum probability
pmin(label) 3. PbestPath/PallPaths
S3 results
37
TMHMM (score 3)
Prediction accuracy vs. coverage
92 bacterial proteins
percent correct
70
45
coverage
Test set bias
38
Experimentally known topologies is a biased
sample
Estimate true performance
39
Correlation between accuracy and TMHMM S3 score
percent correct
mean score
genomes
40
Expected TMHMM performance on proteomes
test set
percent correct
C. elegans
E. coli
S. cerevisiae
coverage
Add C-term.
41
Improved performance by use of experimental
information
92 bacterial proteins
C-terminus known
percent correct
C-terminus not known
75
45
coverage
genomes
42
Expected improvementwhen C-term. location is
known
TMHMMMEMSAT
43
Improved performance by combining TMHMM and MEMSAT
Coverage 60 Accuracy 95
Red C/C Blue F/F White C/F Black F/C
MEMSAT score
TMHMM score
Fusion analysis
44
Experimental topology determination by PhoA
fusions(Manoil, Meth.Cell Biol. 3461)
Periplasmic side
Cytoplasmic side
GFP
45
Experimental topology determination by GFP
fusions(Drew et al., PNAS 992690)
Periplasmic side
Cytoplasmic side
12 topologies
46
GFP and PhoA activitesof 12 E. coli IMPs(PNAS
992690)
GFP
PhoA
YnfA
MarC
PstA
TatC
YaeL
YcbM
YddQ
YdgE
YedZ
YgjV
YiaB
YigG
Medium-scale
47
Medium-scale mapping of E. coli IMPs (N47)(Rapp
et al., in preparation)
results
48
C-in topologies dominate inE. coli
yeast
49
A dual reporter for topology mapping in
yeast(Deak Wolf, JBC 27610663)
50
A dual reporter for topology mapping in
yeast(Deak Wolf, JBC 27610663)
51
Large-scale topology mappingin yeast(Kim von
Heijne, JBC in press)
Glycosylation
Growth on histidinol
YGR290W
YGR055W
YGR105W
YOR376W
-
-
-
-
EndoH
results
52
Results
39 proteins analysed 37 yield consistent results
in the two assays One of the 2 inconsistent
proteins is mitochondrial Predicted C-terminal
location correct for 31 of 37
End