Applied
Soft
Computing
26
(2015)
299–302
Contents
lists
available
at
ScienceDirect
Applied
Soft
Computing
j
ourna
l
h
o
mepage:
www.elsevier.com/locate/asoc
Adaptive
molecular
docking
method
based
on
information
entropy
genetic
algorithm
Zhengfu
Li
a,b,∗
,
Junfeng
Gu
c
,
Hongyan
Zhuang
a
,
Ling
Kang
a
,
Xiaoyu
Zhao
a
,
Quan
Guo
a
a
Department
of
Computer
Science
and
Technology,
Dalian
Neusoft
University
of
Information,
Dalian
116023,
PR
China
b
School
of
Computer
Science
and
Technology,
Dalian
University
of
Technology,
Dalian
116023,
PR
China
c
Department
of
Engineering
Mechanics,
Dalian
University
of
Technology,
Dalian
116023,
PR
China
a
r
t
i
c
l
e
i
n
f
o
Article
history:
Received
26
February
2013
Received
in
revised
form
3
September
2014
Accepted
8
October
2014
Available
online
19
October
2014
Keywords:
Molecular
docking
Genetic
algorithm
Information
entropy
Self-adaptive
Optimization
a
b
s
t
r
a
c
t
Almost
all
the
molecule
docking
models,
using
by
widespread
docking
software,
are
approximate.
Approximation
will
make
the
scoring
function
inaccurate
under
some
circumstances.
This
study
pro-
posed
a
new
molecule
docking
scoring
method:
based
on
force-field
scoring
function,
it
use
information
entropy
genetic
algorithm
to
solve
the
docking
problem.
Empirical-based
and
knowledge-based
sco-
ring
function
are
also
considered
in
this
method.
Instead
of
simple
combination
with
fixed
weights,
coefficients
of
each
factor
are
adaptive
in
the
process
of
searching
optimum
solution.
Genetic
algorithm
with
the
multi-population
evolution
and
entropy-based
searching
technique
with
narrowing
down
space
is
used
to
solve
the
optimization
model
for
molecular
docking
problem.
To
evaluate
this
method,
we
car-
ried
out
a
numerical
experiment
with
134
protein–ligand
complexes
of
the
publicly
available
GOLD
test
set.
The
results
show
that
this
study
improved
the
docking
accuracy
over
the
individual
force-field
sco-
ring
greatly.
Comparing
with
other
popular
docking
software,
it
has
the
best
average
Root-Mean-Square
Deviation
(RMSD).
The
average
computing
time
of
this
study
is
also
good
among
them.
©
2014
Elsevier
B.V.
All
rights
reserved.
1.
Introduction
Molecular
docking
is
to
predict
the
conformation
of
a
ligand
within
the
active
site
of
a
receptor
and
search
for
the
low-energy
binding
modes
[1].
Molecular
dock-
ing
is
widely
used
in
virtual
screen,
and
some
successful
cases
have
been
reported
[2].
The
docking
model
and
scoring
functions
have
received
wide
concerns
in
recent
years
and
a
lot
of
scoring
functions
have
been
proposed
[3].
As
the
core
of
molecular
docking,
scoring
function
can
help
a
docking
program
to
efficiently
explore
the
bind-
ing
space
of
a
ligand.
It
is
also
responsible
for
evaluating
the
binding
affinity
once
the
correct
binding
pose
is
identified
[4].
It
is
an
optimization
process
of
finding
the
best
position
of
a
ligand
in
the
binding
site
of
a
receptor.
A
lot
of
comparative
studies
have
been
done
to
evaluate
the
relative
per-
formances
of
these
widely
used
docking
programs
and
scoring
methods
[5–18].
However,
none
of
these
scoring
functions
or
program
is
generally
applicable
for
all
the
situations
because
the
interactions
between
ligands
and
receptors
are
compli-
cated.
In
addition,
it
is
necessary
to
simplify
docking
models
to
obtain
acceptable
computing
time.
Current
scoring
functions
can
be
roughly
classified
into
three
types:
force
field-
based
scoring
functions,
empirical
scoring
functions
and
knowledge-based
scoring
functions.
These
models
of
widespread
used
docking
functions
are
nearly
approx-
imate
models.
Approximation
makes
one
scoring
function
inaccurate
under
some
circumstances.
Based
on
force-field
scoring
function,
we
also
considered
hydropho-
bic
and
deformation
as
well
in
our
method.
Instead
of
simple
combination
of
them
∗
Corresponding
author
at:
Software
Park
Road
8,
A3-117
Office,
Dalian
116023,
PR
China.
Tel.:
+86
411
84835202;
fax:
+86
411
84835202.
E-mail
address:
lizhengfu@hotmail.com
(Z.
Li).
with
fixed
weights,
coefficients
are
adaptive
in
searching
procedure.
In
order
to
improve
accuracy
and
stability,
knowledge-based
scoring
method
was
used
as
another
scoring
factor
with
adaptive
coefficient.
An
iteration
scheme
in
conjunc-
tion
with
the
multi-population
evolution
and
entropy-based
searching
technique
with
narrowing
down
space
was
used
to
solve
the
optimization
model
for
molec-
ular
docking.
To
evaluate
the
method,
we
performed
the
numerical
experiment
with
134
protein–ligand
complexes
from
the
publicly
available
GOLD
test
set
(
http://www.ccdc.cam.ac.uk/).
The
results
indicated
that
the
scoring
function
for
molecular
docking
had
high
accuracy.
2.
Optimization
model
The
process
of
finding
the
best
pose
is
an
optimization
problem.
The
problem
can
be
described
as
follows:
Min
{F
1
(X)
+
F
2
(X)
+
F
3
(X)}
s.t.
g
i
(X)
<
0,
i
=
1,
2,
.
.
.,
n
(1)
where
X
is
a
vector
of
design
variables,
indicating
the
orientation
and
conformation
information
of
a
ligand.
Due
to
the
computational
reasons,
it
is
always
assumed
that
the
ligand
is
flexible
and
that
the
receptor
is
rigid.
So
X
can
be
defined
as
follows:
X
=
T
x
,
T
y
,
T
z
,
R
x
,
R
y
,
R
z
,
T
b1
,
T
b2
,
.
.
.,
T
bn
,
C
1
,
C
2
,
C
3
T
(2)
http://dx.doi.org/10.1016/j.asoc.2014.10.008
1568-4946/©
2014
Elsevier
B.V.
All
rights
reserved.