accurate protein-protein docking with rapid calculation
Post on 29-Nov-2014
1.696 Views
Preview:
DESCRIPTION
TRANSCRIPT
Improvement of the Protein-Protein Docking Prediction
by Introducing a Simple Hydrophobic Interaction Model:
an Application to Interaction Pathway Analysis
Masahito Ohue† ‡ Yuri Matsuzaki† Takashi Ishida† Yutaka Akiyama†
† Graduate School of Information Science and Engineering,
Tokyo Institute of Technology
‡ JSPS Research Fellow
The 7th IAPR International Conference on
Pattern Recognition in Bioinformatics
November 9th, 2012, Tokyo, Japan
• MEGADOCK is a PPI network prediction system
using all-to-all protein docking calculation
– Required a rapidly calculation for all-to-all prediction
– MEGADOCK is 8.8 times faster than ZDOCK
• We proposed new docking score model
added a hydrophobic interaction keeping rapidness
• We achieved the better level of accuracy without
increasing calculation time
2
Summary
2012/11/09 PRIB2012
Introduction
• Proteins play a key role in biological events
Many proteins display their biological functions
by binding to a specific partner protein at a specific site
4
Background
2012/11/09 PRIB2012
ex) Signal transduction
SOS
Ras Raf1
http://en.wikipedia.org/wiki/Epidermal_growth_factor_receptor
• PDB crystal structures are currently increasing
but interacted structures are not enough
5
Background
2012/11/09 PRIB2012
Structural coverage of interactions
While there is structural data for a large part of the interactors (60–80%), the portion of interactions
having an experimental structure or a template for comparative modeling is limited (less than 30%)
→ Computational protein docking is very important
Modified from Stein A, 2011.
• Protein docking is useful of all-to-all protein-
protein interaction prediction
6
Background
2012/11/09 PRIB2012
Need a fast protein docking software for all-to-all
protein-protein interaction prediction
• 1000 proteins’ combinations
→500,500 (1000x1001/2)
• 500,500 PPI predictions require
43,000 CPU days when using ZDOCK*
• On 400 nodes x 12 cores CPU/node
→ still require 9 days
*Mintseris J, et al. Proteins, 2007.
• MEGADOCK has a rapid protein docking engine
• However, MEGADOCK’s docking accuracy is
worse compared to ZDOCKs
• Our purpose is developing more accurate protein
docking method with rapidness 7
Background
2012/11/09 PRIB2012
0
30
60
90
120
150
MEGADOCK ZDOCK 2.3 ZDOCK 3.0
Avg
do
ckin
g t
ime [
min
]
All-to-all PPI Prediction
in 1000 proteins
9 days → 1 day ZDOCK 3.0 MEGADOCK
8
Protein Docking Problem
2012/11/09 PRIB2012
Protein monomeric
structure pair
Docking calculation
Protein complex
(predicted structure)
9
MEGADOCK Docking Engine Overview
2012/11/09 PRIB2012
Receptor FFT
Ligand FFT
Loop for Ligand rotations
Receptor voxelization
Convolution
Score calculation (Inverse FFT)
Save high scoring docking poses O
pe
nM
P P
arallelizatio
n
Ligand voxelization
…
…
Receptor
Ligand Master
Node
Worker
Node 1
Worker
Node 2
Worker
Node 3
MPI Parallelization
• 3-D Grid Convolution
10
3-D Grid Convolution by FFT
2012/11/09 PRIB2012
Katchalski-Katzir E, et al. PNAS, 1992.
Receptor Ligand
Convolution using FFT
Convolution
Docking Score Parallel translation
Δθ is 15 degree → 3600
patterns of Ligand rotations
(2-D image)
Previous Score Model (MEGADOCK)
• MEGADOCK’s docking score
– Calculate shape complementarity and electrostatics
with only 1 convolution
rPSC ELEC
11
Ohue M, et al. IPSJ TOM, 2010.
2012/11/09 PRIB2012
Convolution
rPSC
• rPSC (real Pairwise Shape Complementarity)
– Shape complementarity score using only real value
Ohue M, et al. IPSJ TOM, 2010.
12 2012/11/09 PRIB2012
2 -27 -27 -27 -27 -27 2
3 -27 -27 -27 -27 -27 2
3 -27 -27 -27 5 2
3 -27 -27 -27 5 2
3 -27 -27 -27 -27 -27 2
2 -27 -27 -27 -27 -27 2
2 3 3 3 2
2 3 3 3 2
1 1
1 1
1 1
2 2
2 2
1 1
1
1
Receptor rPSC Ligand rPSC
1 1
1
1
11
Receptor
Ligand
Good point!
Penalty
(2-D image)
1 1
1 1
1 1
1 1
1 1
1 1
1
1
• ZDOCK 3.0 convolution
• MEGADOCK convolution
13
Comparison of Convolutions
2012/11/09 PRIB2012
Shape Complementarity = Fit neatly Clash +i Electrostatics = Coulomb force
Desolvation free energy 1 = atom a’s effect atom b’s effect +i Desolvation free energy 2
・・・
= atom c’s effect atom d’s effect +i
Desolvation free energy 6 = atom k’s effect atom l’s effect +i
Docking Score Shape Complementarity (rPSC) +i = Coulomb force
Mintseris J, et al. Proteins, 2007.
Accurate and Fast Docking
• For more accurate protein docking
– Considering hydrophobic interaction
– Previous methods require high calculation cost
• ZDOCK 2.3・・・2 convolutions
• ZDOCK 3.0・・・8 convolutions
• For keeping rapid calculation
– We should keep 1 time convolution
• We proposed simple hydrophobic interaction
model with keeping 1 convolution
14 2012/11/09 PRIB2012
Materials and Methods
• Implementation of hydrophobic interaction
– Used Atomic Contact Energy score
• Atomic Contact Energy (ACE)
– The empirical atomic pairwise potential for estimating
desolvation free energies
– Used by ZDOCK, FireDock*, etc.
• How to add the ACE to MEGADOCK?
– Add the averaged ACE value (non-pairwise) of surface
atom of Receptor
16
Proposed Model
2012/11/09 PRIB2012
Zhang C, et al. J Mol Biol, 1997.
*Andrusier N, et al. Proteins, 2007.
Modified rPSC Model
Modification of rPSC Model
• Considered hydrophobic surface of the receptor
using non-pairewise ACE
17 2012/11/09 PRIB2012
Sum of near atoms’ ACE (space)
(inside of the receptor)
2+H -27 -27 -27 -27 -27 2+H
3+H -27 -27 -27 -27 -27 2+H
3+H -27 -27 -27 5+H 2+H
3+H -27 -27 -27 5+H 2+H
3+H -27 -27 -27 -27 -27 2+H
2+H -27 -27 -27 -27 -27 2+H
2+H 3+H 3+H 3+H 2+H
2+H 3+H 3+H 3+H 2+H
1 1
1 1
1 1
1 1
1 1
1 1
1
1
1+H 1+H
1+H
1+H
1+H 1+H
Receptor
Ligand
Pairwise Potential
• Using table of contact energy
18
eij N Cα C …
N -0.724 -0.903 -0.722 …
Cα -0.119 -0.842 -0.850 …
C -0.008 -0.077 -0.704 …
… … … … …
Receptor
eiN
Ligand
Set 1 on
position of
N atoms
×
FFT
Receptor
eiCα
Ligand
Set 1 on
position of
Cα atoms
×
FFT
Receptor
eiC
Ligand
Set 1 on
position of
C atoms
×
FFT ←Need (#type of atoms/2) times convolutions
・・・
Averaged (Non-Pairwise) Potential
• Using averaged contact energy
19
eij N Cα C … ei
N -0.724 -0.903 -0.722 … -0.495
Cα -0.119 -0.842 -0.850 … -0.553
C -0.008 -0.077 -0.704 … -0.464
… … … … … … ↑ 1 time FFT
Receptor
ei
Ligand
Set 1 on
position of
all atoms
×
FFT
Pros : high speed calculation
Cons: poor accuracy
• Dataset: Protein-protein docking benchmark 4.0
– 176 complex structures
(include both bound/unbound form)
20
Experimental Settings
2012/11/09 PRIB2012
Hwang H, et al. Proteins, 2010.
PDB ID: 1CGI
1CGI receptor
1CGI ligand
2CGA
1HPT
Predicted 1CGI
using 2CGA and 1HPT
Predicted 1CGI
using 1CGI_r and 1CGI_l
bound
unbound
(re-docking)
(real problem)
Proposed Docking Score
• New docking score
– Calculate 3 physico-chemical effects
by only 1 convolution calculation
21 2012/11/09 PRIB2012
• How to evaluate accuracy? → Success Rank
22
Experimental Settings
2012/11/09 PRIB2012
Native
S=2132.1 35.1Å
S=1802.3 2.3Å
S=1623.5 4.1Å
S=300.8 24.8Å
・・・
1st
2nd
3rd
3600th
Docking score Ligand-RMSD
・・・
Success Rank
of this case is 2
Near native solution
(L-RMSD < 5Å)
Ligand-RMSD (root mean square deviation) RMSD with respect to the X-ray structure, calculated for the all heavy atoms of the ligand residues
when only the receptor proteins (predicted structure and X-ray structure) are superimposed.
Results and Discussions
0
100
200
300
400
MEGADOCK Proposed ZDOCK 2.3 ZDOCK 3.0
To
tal ti
me
fo
r 1
76
do
ck
ing
[h
r]
Docking Calculation Time
24
good
2012/11/09 PRIB2012
On TSUBAME 2.0, 1 CPU core (Intel Xeon 2.93 GHz)
Proposed
Comparable
x8.8 faster than ZDOCK 3.0 / x3.8 faster than ZDOCK 2.3
Accuracy in Bound Set
25
good
2012/11/09 PRIB2012
0%
20%
40%
60%
80%
100%
1 10 100 1000
Success R
ate
Number of Predictions
ZDOCK 3.0
ZDOCK 2.3
Proposed
MEGADOCK
ex) The number of cases of “Success Rank ≦ 100”
is 147 (84%(147/176) of all cases).
Achieved the same level of accuracy as ZDOCK 3.0
26
Accuracy in Unbound Set
2012/11/09 PRIB2012
0%
20%
40%
60%
80%
100%
1 10 100 1000
Success R
ate
Number of Predictions
ZDOCK 3.0
ZDOCK 2.3
Proposed
MEGADOCK
ex) The number of cases of “Success Rank ≦ 1000”
is 46 (26%(46/176) of all cases).
good
Achieved the same level of accuracy as ZDOCK 2.3
Prediction Example
27
PDB ID: 1BVN (Unbound Success Rank:
11(MEGADOCK)→1(Proposed))
Receptor: α-amylase
Ligand: Tendamistat
(α-amylase inhibitor)
2012/11/09 PRIB2012
Hydrophobicity scale*
hydrophilic hydrophobic
*Eisenberg D, et al. J Mol Biol, 1984.
The 1st rank predicted structure
by MEGADOCK (wrong position)
Another angle
• Bacterial chemotaxis signaling pathway prediction
– Predicting “interact or not” from docking results*
• Method
28
Application to Pathway Analysis
2012/11/09 PRIB2012
* Matsuzaki Y, et al. JBCB, 2009.
S1
S2
S3
・・・
S6,000
Mean of S μ
S.D. of S σ
Docking
calculationZR2
ZR3,273
ZR1
・・・
・・・
ZRANKRank of
energy score: ZR
• Bacterial chemotaxis signaling pathway prediction
– Predicting “interact or not” from docking results*
– Results (F-measure)
29
Application to Pathway Analysis
2012/11/09 PRIB2012
* Matsuzaki Y, et al. JBCB, 2009. d A(L) B R W D Y C Z X FliG FliM FliN
Tsr ◆ ◆ ◆ ◆ ◆
A(L) ◆ ◆ ◆ ◆ ◆ ◆ ◆ ◆ Known PPI
B ◆ ◆ ◆ ◆ ◆ Predicted
R ◆ ◆
W ◆ ◆ ◆ ◆ ◆ ◆
D ◆ ◆ ◆ ◆
Y ◆ ◆ ◆ ◆ ◆ ◆ ◆ ◆ ◆ ◆ ◆
C ◆ ◆ ◆ ◆ ◆
Z ◆ ◆ ◆ ◆
X ◆ ◆ ◆ ◆ ◆ ◆
FliG ◆ ◆ ◆ ◆ ◆
FliM ◆
FliN ◆ ◆ ◆
MEGADOCK Proposed ZDOCK 3.0
0.43 0.45 0.49
Conclusion
Conclusion
• We proposed novel docking score including
simple hydrophobic interaction
• We achieved the better level of accuracy without
increasing calculation time
– Same level of accuracy as ZDOCK 2.3
– 8.8 times faster than ZDOCK 3.0
3.8 times faster than ZDOCK 2.3
• Future works
– Developing a hydrophobic interaction model considering
both receptor and ligand protein
– Applying to large PPI network and cross-docking of
structure ensemble
31 2012/11/09 PRIB2012
• Akiyama Lab. Members
• This work was supported in part by
– a Grant-in-Aid for JSPS Fellows, 23・8750
– a Grant-in-Aid for Research and Development of The Next-Generation
Integrated Life Simulation Software
– Educational Academy of Computational Life Sciences
from the Ministry of Education, Culture, Sports, Science and
Technology of Japan (MEXT).
Acknowledgements
32
TECH chan
top related