I
AE
S
I
nte
rna
t
io
na
l J
o
urna
l o
f
Art
if
icia
l In
t
ellig
ence
(
I
J
-
AI
)
Vo
l.
3
,
No
.
4
,
Dec
em
b
er
201
4
,
p
p
.
150
~
155
I
SS
N:
2252
-
8938
150
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ia
e
s
jo
u
r
n
a
l.c
o
m/o
n
lin
e/in
d
ex
.
p
h
p
/I
J
A
I
Qu
an
tile Reg
ressio
n
Neural N
etwo
rks
Based
Pr
ed
ictio
n
o
f
Dr
u
g
Activ
ities
M
o
ha
m
m
e
d E
.
E
l
-
T
elba
ny
Co
m
p
u
ters
a
n
d
S
y
ste
m
s De
p
a
rt
m
e
n
t
El
e
c
tro
n
ics
Re
se
a
rc
h
In
stit
u
te
,
Eg
y
p
t
Art
icle
I
nfo
AB
ST
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
J
u
l
3
,
2
0
1
4
R
ev
i
s
ed
No
v
9
,
2
0
1
4
A
cc
ep
ted
No
v
2
1
,
2
0
1
4
QSA
R
(q
u
a
n
ti
tativ
e
stru
c
tu
re
-
a
c
ti
v
it
y
re
latio
n
sh
ip
)
m
o
d
e
li
n
g
is
o
n
e
o
f
th
e
w
e
ll
d
e
v
e
lo
p
e
d
a
re
a
s
in
d
ru
g
d
e
v
e
lo
p
m
e
n
t
th
r
o
u
g
h
c
o
m
p
u
tatio
n
a
l
c
h
e
m
istr
y
.
S
im
il
a
r
m
o
lec
u
les
w
it
h
ju
st
a
sli
g
h
t
v
a
riatio
n
in
th
e
ir
str
u
c
tu
re
c
a
n
h
a
v
e
q
u
it
d
if
fe
re
n
t
b
io
lo
g
ica
l
a
c
ti
v
i
ty
.
T
h
is
k
in
d
o
f
re
latio
n
sh
i
p
b
e
tw
e
e
n
m
o
lec
u
lar
stru
c
tu
re
a
n
d
c
h
a
n
g
e
in
b
io
l
o
g
ica
l
a
c
ti
v
it
y
is
c
e
n
ter
o
f
f
o
c
u
s
f
o
r
QS
A
R
M
o
d
e
li
n
g
.
P
re
d
icti
o
n
s
o
f
p
ro
p
e
rt
y
a
n
d
/o
r
a
c
ti
v
it
y
o
f
in
tere
st
h
a
v
e
th
e
p
o
ten
ti
a
l
t
o
sa
v
e
ti
m
e
,
m
o
n
e
y
a
n
d
m
in
i
m
ize
th
e
u
se
o
f
e
x
p
e
n
siv
e
e
x
p
e
ri
m
e
n
tal
d
e
sig
n
s,
su
c
h
a
s,
f
o
r
e
x
a
m
p
le,
a
n
i
m
a
l
tes
ti
n
g
.
In
telli
g
e
n
t
m
a
c
h
in
e
lea
rn
in
g
tec
h
n
i
q
u
e
s
a
re
im
p
o
rtan
t
to
o
ls
f
o
r
QSA
R
a
n
a
ly
sis,
a
s
a
re
su
lt
,
th
e
y
a
r
e
in
teg
ra
ted
in
to
th
e
d
ru
g
p
ro
d
u
c
ti
o
n
p
ro
c
e
ss
.
T
h
e
e
ff
e
c
ti
v
e
lea
rn
a
b
le
m
o
d
e
l
c
a
n
re
d
u
c
e
t
h
e
c
o
st
o
f
d
ru
g
d
e
sig
n
sig
n
if
ica
n
tl
y
.
T
h
e
q
u
a
n
t
il
e
e
stim
a
ti
o
n
v
ia
n
e
u
ra
l
n
e
t
w
o
rk
stru
c
tu
re
tec
h
n
i
q
u
e
in
tr
o
d
u
c
e
d
in
th
is
p
a
p
e
r
is
u
se
d
t
o
p
re
d
ict
a
c
ti
v
it
y
o
f
p
y
ri
m
id
in
e
s
b
a
se
d
o
n
t
h
e
stru
c
tu
re
–
a
c
ti
v
it
y
re
latio
n
sh
ip
o
f
th
e
se
c
o
m
p
o
u
n
d
s
w
h
ich
a
ss
ist
f
o
r
f
in
d
in
g
p
o
ten
ti
a
l
trea
t
m
e
n
t
a
g
e
n
ts
f
o
r
se
rio
u
s
d
ise
a
se
.
In
c
o
m
p
a
riso
n
w
it
h
sta
ti
stica
l
q
u
a
n
ti
le
re
g
re
ss
io
n
,
th
e
q
r
n
n
sig
n
if
ica
n
tl
y
re
d
u
c
e
th
e
p
re
d
ictio
n
e
rro
r
.
K
ey
w
o
r
d
:
Ma
ch
i
n
e
L
ea
r
n
i
n
g
P
r
ed
ictio
n
QS
A
R
Q
u
a
n
tile
Ne
u
r
al
Net
w
o
r
k
s
Co
p
y
rig
h
t
©
2
0
1
4
In
stit
u
te o
f
A
d
v
a
n
c
e
d
E
n
g
i
n
e
e
rin
g
a
n
d
S
c
ien
c
e
.
Al
l
rig
h
ts
re
se
rv
e
d
.
C
o
r
r
e
s
p
o
nd
ing
A
uth
o
r
:
Mo
h
a
m
m
ed
E
.
E
l
-
T
elb
an
y
,
Dep
ar
te
m
en
t o
f
E
lectr
ical
an
d
C
o
m
p
u
ter
E
n
g
in
ee
r
i
n
g
,
C
o
m
p
u
ter
s
an
d
S
y
s
te
m
s
Dep
ar
t
m
en
t E
lectr
o
n
ics
R
esear
ch
I
n
s
ti
tu
te,
E
g
y
p
t
.
E
m
ail:
telb
a
n
y
@
er
i.sci.e
g
1.
I
NT
RO
D
UCT
I
O
N
I
n
r
ec
en
t y
ea
r
s
,
w
ith
p
r
o
d
u
cts
o
f
h
u
m
an
g
en
o
m
e
p
r
o
j
e
ct
h
e
lp
in
g
to
r
ev
ea
l
m
an
y
n
e
w
d
is
ea
s
e
ta
r
g
e
ts
t
o
w
h
ich
d
r
u
g
tr
e
atm
en
ts
m
ig
h
t
b
e
aim
ed
,
a
ll
th
e
m
ajo
r
p
h
ar
m
ac
eu
t
ica
l
c
o
m
p
an
ies
h
av
e
in
v
este
d
h
ea
v
i
ly
in
th
e
r
o
u
t
in
e
u
lt
r
a
-
Hig
h
T
h
r
o
u
g
h
p
u
t
Scr
ee
n
in
g
(
u
H
T
S)
o
f
v
as
t
n
u
m
b
er
s
o
f
„
d
r
u
g
-
l
ik
e
‟
m
o
le
cu
les
g
u
i
d
e
d
b
y
ch
em
o
in
f
o
r
m
atics
in
v
es
tig
ati
o
n
s
[
L
ip
in
s
k
i
,
2
0
0
4
,
L
ee
s
o
n
el
at.
,
2
0
0
4
]
.
T
h
e
d
ev
e
l
o
p
m
en
t
o
f
a
n
ew
d
r
u
g
is
s
til
l
a
ch
a
llen
g
in
g
,
tim
e
-
co
n
s
u
m
in
g
an
d
c
o
s
t
-
in
ten
s
iv
e
p
r
o
c
es
s
an
d
d
u
e
t
o
th
e
en
o
r
m
o
u
s
ex
p
en
s
e
o
f
f
ailu
r
es
o
f
ca
n
d
id
ate
d
r
u
g
s
lat
e
in
th
eir
d
ev
e
lo
p
m
en
t,
u
H
T
S
in
v
itr
o
ass
ay
s
n
o
w
co
v
er
li
a
b
ili
ties
s
u
ch
as
p
o
s
s
i
b
l
e
s
id
e
ef
f
ec
ts
[
L
i,
2
0
0
5
]
as
w
ell
as
th
er
ap
eu
tic
p
r
o
p
er
ties
,
c
o
m
p
u
tat
io
n
al
m
eth
o
d
s
ca
n
b
e
u
s
e
d
t
o
ass
is
t
a
n
d
s
p
e
ed
u
p
th
e
d
r
u
g
d
esig
n
p
r
o
ce
s
s
.
I
t
is
o
b
v
io
u
s
th
at
th
e
d
r
u
g
d
is
co
v
e
r
y
an
d
d
ev
el
o
p
m
en
t
p
r
o
c
ess
w
o
u
l
d
g
r
ea
t
ly
b
en
ef
it
f
r
o
m
f
aster
an
d
ch
ea
p
e
r
p
r
o
ce
d
u
r
es
t
o
id
en
tif
y
ch
em
ical
c
o
m
p
o
u
n
d
s
w
ith
d
esi
r
e
d
b
i
o
l
o
g
i
ca
l
p
r
o
p
er
ties
an
d
t
o
o
p
tim
ize
th
ei
r
s
t
r
u
c
tu
r
e
in
o
r
d
er
t
o
o
b
ta
in
ef
f
ec
t
iv
e
d
r
u
g
s
.
Se
v
er
a
l
m
ajo
r
b
o
tt
len
e
ck
s
in
d
r
u
g
d
is
c
o
v
e
r
y
m
a
y
b
e
ad
d
r
ess
e
d
w
ith
co
m
p
u
ter
-
ass
i
s
ted
d
r
u
g
d
es
ig
n
m
eth
o
d
s
,
s
u
ch
as
q
u
an
t
ita
tiv
e
s
tr
u
ctu
r
e
–
a
ctiv
ity
r
elat
io
n
s
h
i
p
s
(
QSA
R
)
m
o
d
els
[
Han
s
ch
,
1
9
6
9
]
,
w
h
er
e
th
e
m
o
lecu
la
r
ac
t
iv
ities
a
r
e
c
r
it
ica
l
f
o
r
d
r
u
g
d
esig
n
,
th
ey
ca
n
b
e
p
r
e
d
ic
te
d
b
y
QSA
R
m
o
d
els
.
Mo
lecu
lar
ac
ti
v
it
y
i
s
d
eter
m
i
n
ed
b
y
its
s
tr
u
ctu
r
e,
s
o
s
tr
u
ct
u
r
e
p
ar
a
m
eter
s
ar
e
e
x
tr
ac
ted
b
y
d
if
f
er
en
t
m
et
h
o
d
s
to
b
u
ild
Q
S
A
R
m
o
d
els.
No
w
ad
a
y
s
,
m
ac
h
in
e
lear
n
in
g
alg
o
r
ith
m
s
h
av
e
b
ee
n
u
s
e
d
in
th
e
m
o
d
elin
g
o
f
QSA
R
p
r
o
b
lem
s
[
Du
ch
et
al.
,
2
0
0
7
,
C
h
in
a
n
d
C
h
u
n
,
2
0
1
2
;
Ger
tr
u
d
esa
et
al.
,
2
0
1
2
]
.
T
h
ey
ex
tr
ac
t
in
f
o
r
m
atio
n
f
r
o
m
ex
p
e
r
im
en
tal
d
at
a
b
y
co
m
p
u
tati
o
n
al
an
d
s
tat
is
ti
ca
l
m
eth
o
d
s
an
d
g
en
e
r
a
te
a
s
et
o
f
r
u
les
,
f
u
n
cti
o
n
s
o
r
p
r
o
ce
d
u
r
es
th
a
t
al
lo
w
th
em
t
o
p
r
e
d
i
ct
th
e
p
r
o
p
e
r
ti
es
o
f
n
o
v
e
l
o
b
jects
th
at
a
r
e
n
o
t
in
clu
d
e
d
in
th
e
l
ea
r
n
in
g
s
et
.
Evaluation Warning : The document was created with Spire.PDF for Python.
IJ
-
AI
I
SS
N:
2252
-
8938
Qu
a
n
tile R
eg
r
ess
io
n
N
eu
r
a
l N
etw
o
r
ks B
a
s
ed
P
r
ed
ictio
n
o
f D
r
u
g
A
ctivities
(
Mo
h
a
mme
d
E
.
E
l
-
Telb
a
n
y
)
151
Fo
r
m
ally
,
a
lea
r
n
in
g
alg
o
r
i
th
m
is
task
ed
w
ith
s
elec
tin
g
a
h
y
p
o
th
esis
th
at
b
es
t
s
u
p
p
o
r
ts
th
e
d
a
ta
.
C
o
n
s
i
d
e
r
in
g
th
e
h
y
p
o
th
esis
t
o
b
e
a
f
u
n
cti
o
n
m
ap
p
in
g
f
r
o
m
th
e
d
ata
s
p
a
ce
t
o
th
e
r
esp
o
n
s
e
s
p
ac
e
;
i
.
e
.
,
.
T
h
e
lear
n
er
s
elec
ts
th
e
b
est
h
y
p
o
t
h
esi
s
f
r
o
m
a
s
p
ac
e
o
f
all
p
o
s
s
ib
le
h
y
p
o
t
h
ese
s
b
y
m
in
i
m
i
ze
e
r
r
o
r
s
w
h
en
p
r
e
d
ic
tin
g
v
alu
e
f
o
r
n
ew
d
ata,
o
r
if
o
u
r
m
o
d
e
l in
clu
d
es a
c
o
s
t
f
u
n
ctio
n
o
v
e
r
e
r
r
o
r
s
,
t
o
m
in
i
m
ize
th
e
t
o
t
al
c
o
s
t
o
f
er
r
o
r
s
.
Fig
u
r
e
1
.
Gen
er
al
Step
s
o
f
De
v
elo
p
in
g
QS
AR
Mo
d
els.
A
s
s
h
o
w
n
in
Fig
u
r
e
1
,
th
e
QSA
R
m
o
d
elin
g
is
h
ea
v
i
ly
d
ep
en
d
en
t
o
n
th
e
s
ele
cti
o
n
o
f
m
o
lecu
l
ar
d
es
cr
ip
to
r
s
;
if
th
e
ass
o
cia
ti
o
n
o
f
th
e
d
esc
r
i
p
t
o
r
s
s
ele
cte
d
t
o
b
io
l
o
g
ic
al
p
r
o
p
e
r
ty
is
s
tr
o
n
g
th
e
QSA
R
m
o
d
el
ca
n
id
en
t
if
y
v
alid
r
ela
ti
o
n
s
b
etw
ee
n
m
o
lecu
l
ar
f
e
atu
r
es
an
d
b
i
o
lo
g
ic
al
p
r
o
p
e
r
ty
/activ
i
ty
.
T
h
u
s
,
u
n
in
f
o
r
m
ativ
e
o
r
r
e
d
u
n
d
an
t
m
o
lecu
l
a
r
d
es
cr
ip
to
r
s
s
h
o
u
ld
b
e
r
em
o
v
ed
u
s
in
g
s
o
m
e
f
ea
tu
r
e
s
ele
cti
o
n
m
eth
o
d
s
d
u
r
in
g
(
F
i
lter
s
)
o
r
be
f
o
r
e
(
w
r
a
p
p
ers
)
th
e
le
ar
n
in
g
p
r
o
c
ess
.
Su
b
s
eq
u
en
tly
,
f
o
r
tu
n
in
g
an
d
v
alid
ati
o
n
o
f
th
e
p
r
e
d
ic
tiv
ely
o
f
lea
r
n
ed
QSA
R
m
o
d
el,
o
n
e
o
f
th
e
v
alid
ati
o
n
s
tr
a
teg
y
ca
n
b
e
ap
p
li
ed
li
k
es
cr
o
s
s
-
v
ali
d
at
io
n
,
leav
e
-
o
n
e
-
o
u
t
o
r
th
e
f
u
ll
d
a
t
a
s
et
is
d
iv
id
ed
in
t
o
a
t
r
ain
in
g
s
et
an
d
a
t
esti
n
g
s
e
t
p
r
io
r
to
lea
r
n
in
g
(
S
ee
[
E
l
-
T
el
b
an
y
,
2
0
1
4
]
f
o
r
a
s
u
r
v
ey
)
.
A
ctu
ally
,
th
e
m
ac
h
in
e
le
ar
n
in
g
f
ield
[
B
is
h
o
p
,
2
0
0
6
;
Ma
r
s
lan
d
,
2
0
0
9
;
B
u
r
k
e
an
d
Ke
n
d
all;
2
0
1
4
,
Mo
h
r
i,
2
0
1
2
]
h
av
e
v
er
s
at
ile
m
eth
o
d
s
o
r
al
g
o
r
i
th
m
s
s
u
ch
as
d
ec
is
i
o
n
tr
e
es,
la
zy
lear
n
in
g
,
k
-
n
ea
r
e
s
t
n
eig
h
b
o
r
s
,
B
ay
esian
m
eth
o
d
s
,
G
au
s
s
ian
p
r
o
ce
s
s
e
s
,
a
r
tif
i
cia
l
n
eu
r
al
n
etw
o
r
k
s
,
a
r
tif
i
cia
l
im
m
u
n
e
s
y
s
te
m
s
,
p
ar
t
icl
e
-
s
w
ar
m
o
p
tim
izati
o
n
,
a
r
tif
i
cia
l
b
e
e
o
p
t
im
izatio
n
,
cu
ck
o
o
s
ea
r
ch
,
s
u
p
p
o
r
t
v
ec
to
r
m
ac
h
in
es
,
an
d
k
er
n
el
a
lg
o
r
ith
m
s
f
o
r
a
v
ar
i
ety
o
f
tas
k
s
in
d
r
u
g
d
es
ig
n
.
T
h
ese
m
eth
o
d
s
a
r
e
al
te
r
n
ativ
e
s
to
o
b
tain
s
atis
f
y
in
g
m
o
d
e
ls
b
y
tr
ain
in
g
o
n
a
d
a
ta
s
et.
Ho
w
ev
er
,
t
h
e
p
r
ed
ict
io
n
f
r
o
m
m
o
s
t
r
e
g
r
ess
io
n
m
o
d
el
s
-
b
e
i
t
m
u
ltip
le
r
e
g
r
ess
io
n
,
n
e
u
r
al
n
et
w
o
r
k
s
,
s
u
p
p
o
r
t
v
ec
to
r
m
ac
h
i
n
e
(
SVM
)
,
d
ec
is
io
n
tr
ee
s
,
etc.
-
is
a
p
o
in
t
est
i
m
a
te
o
f
t
h
e
co
n
d
itio
n
al
m
ea
n
o
f
a
r
esp
o
n
s
e
(
i.e
.
,
q
u
an
tit
y
b
ei
n
g
p
r
ed
ict
ed
)
,
g
iv
en
a
s
et
o
f
p
r
ed
icto
r
s
.
R
ec
en
t
ad
v
a
n
ce
s
i
n
co
m
p
u
ti
n
g
allo
w
th
e
d
ev
elo
p
m
en
t
o
f
r
eg
r
e
s
s
io
n
m
o
d
el
s
f
o
r
p
r
ed
ictin
g
a
g
i
v
en
q
u
an
t
ile
o
f
th
e
co
n
d
itio
n
al
d
is
tr
ib
u
tio
n
,
b
o
th
p
ar
am
etr
ical
l
y
a
n
d
n
o
n
p
ar
am
etr
icall
y
.
T
h
e
g
e
n
er
al
ap
p
r
o
ac
h
is
ca
lled
Qu
a
n
tile
R
eg
r
ess
io
n
,
b
u
t
t
h
e
m
et
h
o
d
o
lo
g
y
(
o
f
co
n
d
itio
n
al
q
u
an
tile
e
s
ti
m
atio
n
)
ap
p
lies
to
an
y
s
tat
is
tica
l
m
o
d
el,
b
e
it
m
u
ltip
le
r
eg
r
es
s
io
n
,
SVM
v
ec
to
r
m
ac
h
i
n
es,
o
r
r
an
d
o
m
f
o
r
est
s
.
q
u
an
t
ile
r
eg
r
e
s
s
io
n
n
eu
r
al
n
et
w
o
r
k
s
(
QR
N
N)
[
T
ay
lo
r
,
2
0
0
0
]
esti
m
ates
co
n
d
itio
n
al
v
al
u
es
o
f
an
in
d
i
v
id
u
al
q
u
a
n
tile
u
s
i
n
g
a
m
u
ltil
a
y
er
p
er
ce
p
tr
o
n
n
eu
r
al
n
et
w
o
r
k
.
I
n
th
i
s
p
ap
er
,
q
u
an
tile
r
e
g
r
ess
io
n
n
e
u
r
al
n
et
w
o
r
k
s
tec
h
n
iq
u
es
ar
e
u
s
ed
to
s
t
u
d
y
th
e
r
elatio
n
s
h
ip
s
b
et
w
ee
n
d
i
f
f
er
e
n
t
k
n
o
w
n
tar
g
et
ac
ti
v
itie
s
w
ith
r
esp
ec
t
to
th
e
attr
ib
u
tes
o
f
a
f
i
n
ite
n
u
m
b
er
o
f
co
m
p
o
u
n
d
s
.
T
h
e
A
n
eu
r
al
n
et
w
o
r
k
is
u
s
ed
to
esti
m
ate
t
h
e
p
o
ten
ti
all
y
n
o
n
-
li
n
ea
r
q
u
a
n
tile
m
o
d
els
w
h
ich
i
s
m
o
r
e
r
es
is
ta
n
ce
to
t
h
e
o
u
tl
ier
s
.
T
h
e
r
est
o
f
th
e
p
a
p
e
r
is
o
r
g
an
i
ze
d
as
f
o
ll
o
w
s
.
Sect
io
n
2
b
r
i
ef
ly
in
tr
o
d
u
ce
s
th
e
QSA
R
m
o
d
els
an
d
q
u
an
til
e
r
eg
r
ess
i
o
n
.
Secti
o
n
3
is
in
tr
o
d
u
c
es
th
e
q
u
an
tile
n
eu
r
al
n
e
tw
o
r
k
s
.
Se
cti
o
n
4
d
esc
r
i
b
es
th
e
d
ata
s
et
an
d
p
r
e
p
r
o
c
ess
in
g
s
te
p
s
.
Secti
o
n
5
d
esc
r
i
b
es
an
ev
a
l
u
atio
n
o
f
QSA
R
m
o
d
elin
g
an
d
p
r
e
d
ict
i
o
n
r
esu
l
ts
.
S
ec
t
i
o
n
6
d
es
c
r
i
b
es
an
ex
p
e
r
im
en
tal
r
esu
lts
.
Fin
ally
Secti
o
n
7
p
r
esen
ts
th
e
f
in
d
in
g
s
a
n
d
co
n
c
lu
s
io
n
s
.
2.
Q
SA
R
M
O
DE
L
S AN
D
Q
UA
NT
I
L
E
R
E
G
R
E
SS
I
O
N
QSA
R
m
o
d
els
a
r
e
in
ess
en
ce
a
m
ath
em
atica
l
f
u
n
cti
o
n
th
at
r
elat
es
f
e
atu
r
es
an
d
d
es
c
r
i
p
t
o
r
s
g
en
er
at
e
d
f
r
o
m
s
m
all
m
o
lecu
le
s
t
r
u
ctu
r
e
s
to
s
o
m
e
ex
p
e
r
im
en
tal
d
e
te
r
m
in
ed
a
ctiv
ity
o
r
p
r
o
p
e
r
ty
[
L
i
v
i
n
g
s
to
n
e,
1
9
9
5
]
.
T
h
e
s
tr
u
ctu
r
e
-
a
ctiv
i
ty
s
tu
d
y
ca
n
in
d
ic
at
e
w
h
ich
f
ea
tu
r
es
o
f
a
g
iv
e
n
m
o
lecu
l
e
c
o
r
r
e
lat
e
w
ith
its
a
ctiv
ity
,
th
u
s
m
ak
in
g
it
p
o
s
s
i
b
l
e
t
o
s
y
n
th
esize
n
ew
an
d
m
o
r
e
p
o
t
en
t
c
o
m
p
o
u
n
d
s
w
i
th
en
h
an
c
e
d
b
io
lo
g
i
ca
l
ac
t
iv
iti
es.
QSA
R
an
aly
s
is
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8938
IJ
-
AI
Vo
l.
3
,
No
.
4
,
Dec
em
b
er
20
1
4
:
1
5
0
–
1
5
5
152
is
b
ase
d
o
n
th
e
ass
u
m
p
tio
n
th
a
t th
e
b
eh
av
i
o
r
o
f
c
o
m
p
o
u
n
d
s
is
co
r
r
el
at
ed
to
th
e
ch
a
r
ac
te
r
is
t
ic
s
o
f
th
ei
r
s
tr
u
c
tu
r
e.
In
g
en
er
a
l,
a
QSA
R
m
o
d
e
l is
r
ep
r
es
en
te
d
as f
o
ll
o
w
s
:
∑
(
1
)
W
h
e
r
e
th
e
p
a
r
am
eter
s
ar
e
a
s
et
o
f
m
ea
s
u
r
e
d
(
o
r
c
o
m
p
u
ted
)
p
r
o
p
e
r
t
ies
o
f
th
e
co
m
p
o
u
n
d
s
an
d
th
r
o
u
g
h
a
r
e
th
e
c
alcu
l
at
ed
c
o
e
f
f
icien
ts
o
f
th
e
QSA
R
m
o
d
el
.
Qu
an
tile
r
eg
r
ess
i
o
n
w
as
in
tr
o
d
u
ce
d
b
y
Ko
en
k
er
an
d
B
ass
ett
[
1
9
7
8
]
as
a
c
o
m
p
lem
en
t
to
lea
s
t
s
q
u
ar
es
esti
m
atio
n
(
L
S
E
)
o
r
m
ax
im
u
m
lik
elih
o
o
d
est
im
atio
n
(
ML
E
)
an
d
le
a
d
s
to
f
a
r
-
r
ea
ch
in
g
ex
t
en
s
io
n
s
o
f
“
cl
a
s
s
ic
a
l
”
r
eg
r
ess
i
o
n
an
aly
s
is
b
y
esti
m
atin
g
f
am
ilies
o
f
co
n
d
iti
o
n
a
l
q
u
a
n
til
e
s
u
r
fa
ce
s
,
w
h
ich
d
e
s
cr
i
b
e
th
e
r
e
lat
io
n
b
et
w
ee
n
a
o
n
e
-
d
im
en
s
io
n
al
r
esp
o
n
s
e
an
d
a
h
ig
h
d
im
en
s
io
n
al
p
r
e
d
i
ct
o
r
.
Qu
an
til
e
r
eg
r
ess
i
o
n
o
f
f
er
s
a
n
u
m
b
er
o
f
ad
v
an
tag
es
o
v
e
r
le
ast
-
s
q
u
a
r
es
m
eth
o
d
s
.
W
h
il
e
o
r
d
in
a
r
y
least
s
q
u
ar
es
r
eg
r
ess
i
o
n
ty
p
ically
ass
u
m
es
th
at
th
e
e
r
r
o
r
t
er
m
s
ar
e
I
I
D,
n
o
r
m
ally
d
is
t
r
i
b
u
te
d
,
an
d
h
o
m
o
s
ce
d
as
tic
,
q
u
an
ti
le
r
eg
r
ess
i
o
n
d
o
es
n
o
t
r
e
q
u
i
r
e
th
ese
r
est
r
ic
tiv
e
ass
u
m
p
tio
n
s
.
Fu
r
th
er
m
o
r
e,
s
in
c
e
q
u
an
t
ile
r
eg
r
ess
i
o
n
esti
m
ates
q
u
an
til
es
o
f
th
e
co
n
d
iti
o
n
a
l
d
is
t
r
i
b
u
ti
o
n
r
ath
e
r
th
an
th
e
m
ea
n
,
it
is
m
o
r
e
r
esis
tan
t t
o
o
u
tli
e
r
s
th
an
l
ea
s
t
-
s
q
u
a
r
es m
eth
o
d
s
[
Ko
en
k
e
r
,
2
0
0
5
]
.
I
n
co
n
t
r
as
t
t
o
th
e
le
ast
-
s
q
u
a
r
es
l
o
s
s
f
u
n
ctio
n
,
q
u
an
til
e
r
eg
r
ess
i
o
n
m
ak
es
u
s
e
o
f
th
e
asy
m
m
etr
ic
l
o
s
s
f
u
n
ctio
n
as sh
o
w
n
in
Fig
u
r
e
2
.
(
1
)
W
h
e
r
e
co
r
r
es
p
o
n
d
s
to
th
e
q
u
a
n
tile
to
b
e
est
im
ated
[
K
o
en
k
er
,
2
0
0
5
]
.
N
o
te
th
a
t
if
,
i
.
e
.
,
th
e
m
ed
ian
is
b
ein
g
esti
m
ate
d
,
th
e
n
th
is
l
o
s
s
f
u
n
cti
o
n
b
e
co
m
es si
m
p
l
y
.
(
2)
an
d
th
e
s
u
m
o
f
t
h
e
ab
s
o
lu
te
v
alu
e
s
o
f
t
h
e
r
esid
u
als
i
s
m
i
n
i
m
ized
to
p
er
f
o
r
m
r
e
g
r
ess
io
n
.
T
o
f
it
a
m
o
d
el
,
w
e
est
i
m
a
te
u
s
i
n
g
.
̂
∑
(
3
)
W
h
er
e
is
th
e
n
u
m
b
er
o
f
p
ar
am
eter
s
in
o
u
r
m
o
d
el,
s
o
th
a
t
an
d
ar
e
v
ec
to
r
s
o
f
len
g
t
h
.
T
h
is
co
m
p
u
tatio
n
ca
n
n
o
t
b
e
ca
r
r
ied
o
u
t
an
al
y
tical
l
y
,
i
n
co
n
tr
ast
to
th
e
co
m
p
u
tatio
n
o
f
lea
s
t
s
q
u
ar
es
r
eg
r
e
s
s
io
n
.
I
n
s
tead
,
th
i
s
ca
n
b
e
r
ef
o
r
m
u
lat
ed
as a
p
r
o
b
lem
i
n
li
n
ea
r
p
r
o
g
r
a
m
m
in
g
[
Ko
en
k
er
,
2
0
0
5
]
.
Fig
u
r
e
2
.
T
h
e
q
u
an
tile r
eg
r
ess
i
o
n
lo
s
s
f
u
n
ctio
n
.
3.
Q
UAN
T
I
L
E
NE
U
RAL N
E
T
WO
RK
S
A
r
tif
icia
l
n
eu
r
al
n
etw
o
r
k
s
is
o
n
e
o
f
m
ac
h
in
e
l
ea
r
n
in
g
t
ec
h
n
iq
u
es
w
h
ich
h
av
e
b
ee
n
d
e
v
elo
p
e
d
as
g
en
er
a
liz
ati
o
n
s
o
f
m
ath
em
atica
l
m
o
d
els
o
f
b
io
lo
g
i
ca
l
n
er
v
o
u
s
s
y
s
tem
s
.
T
h
e
le
a
r
n
in
g
ca
p
a
b
il
ity
o
f
an
ar
t
if
ici
al
n
eu
r
o
n
is
ac
h
i
ev
e
d
b
y
ad
j
u
s
ti
n
g
th
e
w
eig
h
ts
in
ac
c
o
r
d
an
c
e
to
th
e
ch
o
s
en
l
ea
r
n
in
g
alg
o
r
ith
m
.
T
h
e
le
ar
n
in
g
s
itu
ati
o
n
s
in
n
eu
r
al
n
etw
o
r
k
s
m
a
y
b
e
class
if
ie
d
in
t
o
th
r
e
e
d
is
t
in
ct
s
o
r
ts
.
T
h
ese
a
r
e
s
u
p
e
r
v
is
e
d
lea
r
n
in
g
,
u
n
s
u
p
er
v
is
e
d
lea
r
n
in
g
an
d
r
ein
f
o
r
ce
m
en
t
lea
r
n
in
g
[
1
2
]
.
T
h
e
m
o
s
t
w
id
ely
-
u
s
ed
n
eu
r
al
n
etw
o
r
k
f
o
r
p
r
e
d
i
cti
o
n
is
Evaluation Warning : The document was created with Spire.PDF for Python.
IJ
-
AI
I
SS
N:
2252
-
8938
Qu
a
n
tile R
eg
r
ess
io
n
N
eu
r
a
l N
etw
o
r
ks B
a
s
ed
P
r
ed
ictio
n
o
f D
r
u
g
A
ctivities
(
Mo
h
a
mme
d
E
.
E
l
-
Telb
a
n
y
)
153
th
e
s
in
g
le
h
i
d
d
en
l
ay
er
f
e
ed
-
f
o
r
w
ar
d
n
etw
o
r
k
.
I
t
co
n
s
is
ts
o
f
a
s
et
o
f
in
p
u
ts
,
w
h
ich
ar
e
c
o
n
n
ec
te
d
to
e
ac
h
o
f
u
n
its
in
a
s
in
g
le
h
i
d
d
en
l
ay
er
,
w
h
ich
,
in
tu
r
n
,
a
r
e
c
o
n
n
ec
te
d
t
o
an
o
u
t
p
u
t
(
s
e
e
F
ig
u
r
e
3
)
.
Fig
u
r
e
3
.
Stru
ct
u
r
e
o
f
th
e
n
eu
r
al
n
et
w
o
r
k
.
T
h
e
r
esu
l
tan
t
m
o
d
el
c
an
b
e
w
r
itten
as
W
h
e
r
e
an
d
ar
e
a
ctiv
ati
o
n
f
u
n
cti
o
n
s
,
w
h
ich
ar
e
f
r
e
q
u
en
tly
c
h
o
s
en
as
s
ig
m
o
id
al
an
d
lin
e
ar
r
es
p
e
ctiv
e
ly
,
an
d
an
d
ar
e
th
e
w
eig
h
ts
(
p
a
r
am
eter
s
)
t
o
b
e
es
tim
ated
[
T
ay
lo
r
2
0
0
0
]
.
T
h
e
p
ar
am
eter
s
o
f
th
e
n
etw
o
r
k
(
i
.
e
.
w
eig
h
ts
)
a
r
e
est
im
ated
b
y
o
p
t
im
izin
g
an
o
b
je
ctiv
e
f
u
n
cti
o
n
(
e
.
g
.
,
v
i
a
m
in
im
izin
g
least
s
q
u
ar
e
er
r
o
r
)
.
I
n
s
t
ea
d
o
f
f
itti
n
g
a
l
in
ea
r
q
u
an
tile
f
u
n
ctio
n
u
s
in
g
th
e
ex
p
r
ess
i
o
n
in
(
5
)
,
a
q
u
an
til
e
r
eg
r
ess
io
n
n
eu
r
a
l
n
etw
o
r
k
m
o
d
el
,
,
o
f
th
e
q
u
an
ti
le
ca
n
b
e
esti
m
ate
d
u
s
in
g
th
e
f
o
ll
o
w
in
g
m
in
i
m
izatio
n
.
(
∑
∑
∑
∑
)
(
5
)
W
h
er
e
an
d
ar
e
r
eg
u
lar
izatio
n
p
ar
a
m
eter
s
w
h
ich
p
e
n
alis
e
th
e
co
m
p
le
x
it
y
o
f
th
e
n
et
w
o
r
k
an
d
th
u
s
av
o
id
o
v
er
f
itti
n
g
[
B
is
h
o
p
,
1
9
9
6
;
2
0
0
6
]
.
T
h
e
o
p
tim
a
l
v
a
lu
es
o
f
w
h
er
e
an
d
an
d
th
e
n
u
m
b
er
,
,
o
f
u
n
i
ts
i
n
th
e
h
id
d
en
la
y
er
ca
n
b
e
estab
lis
h
ed
b
y
cr
o
s
s
-
v
alid
ati
o
n
[
B
is
h
o
p
,
1
9
9
6
; 2
0
0
6
]
.
4.
DATA S
E
T
AND
P
RE
P
RO
CE
SS
I
N
G
T
h
e
d
atas
ets
u
s
ed
in
th
is
s
tu
d
y
ar
e
o
b
t
ain
e
d
f
r
o
m
th
e
UC
I
Data
R
e
p
o
s
it
o
r
y
[
New
m
a
n
et
a
l
.
,
1
9
9
8
]
.
P
y
r
im
id
in
es
d
a
tas
et
c
o
n
tain
s
7
4
d
r
u
g
s
,
an
d
e
ac
h
d
r
u
g
h
as
th
r
ee
p
o
s
s
i
b
l
e
s
u
b
s
titu
t
io
n
p
o
s
i
ti
o
n
s
.
E
ac
h
s
u
b
s
ti
tu
en
t
is
ch
a
r
ac
te
r
i
ze
d
b
y
9
ch
em
ical
p
r
o
p
e
r
t
ies
f
e
atu
r
es:
p
o
la
r
ity
,
s
ize
,
f
lex
i
b
il
ity
,
h
y
d
r
o
g
en
-
b
o
n
d
d
o
n
o
r
,
h
y
d
r
o
g
en
-
b
o
n
d
a
cc
ep
to
r
,
d
o
n
o
r
,
ac
c
ep
t
o
r
,
p
o
la
r
i
z
a
b
i
lity
an
d
ef
f
ec
t.
D
r
u
g
ac
tiv
iti
es
a
r
e
i
d
e
n
tif
ied
b
y
th
e
s
u
b
s
titu
en
ts
.
T
h
e
P
y
r
im
id
in
es
d
at
ase
t
is
r
an
d
o
m
ly
s
h
u
f
f
led
a
n
d
s
p
lit
in
t
o
2
p
a
r
ts
in
th
e
p
r
o
p
o
r
t
io
n
o
f
2
:
1
.
On
e
p
a
r
t
is
u
s
e
d
as
th
e
t
r
a
in
in
g
s
et
,
w
h
ich
co
n
ta
in
s
p
ai
r
s
o
f
5
2
co
m
p
o
u
n
d
s
.
T
h
e
o
th
e
r
p
a
r
t
is
ch
o
s
en
as
th
e
u
n
s
e
en
test
in
g
s
et
,
w
h
ich
co
n
tain
s
p
air
s
o
f
th
e
lef
t
2
2
co
m
p
o
u
n
d
s
an
d
th
o
s
e
b
etw
ee
n
th
e
2
2
co
m
p
o
u
n
d
s
an
d
th
e
tr
a
in
in
g
5
2
co
m
p
o
u
n
d
s
.
Du
e
t
o
th
e
“
cu
r
s
e
o
f
d
i
me
n
s
i
o
n
a
l
ity
”
p
r
o
b
lem
,
s
ea
r
ch
in
g
f
o
r
in
f
o
r
m
ati
v
e
co
m
p
o
u
n
d
s
s
et
as
a
p
r
ep
r
o
ce
s
s
in
g
s
tep
p
r
io
r
to
th
e
ap
p
licatio
n
o
f
q
r
n
n
alg
o
r
ith
m
is
i
m
p
o
r
tan
t
f
o
r
m
an
y
r
ea
s
o
n
s
.
O
n
e
r
ea
s
o
n
is
,
th
at
t
h
e
p
r
ed
ictio
n
ac
cu
r
ac
y
o
f
th
e
q
r
n
n
d
ec
r
ea
s
es
w
h
e
n
ir
r
elev
an
t
o
r
r
ad
ian
t
f
ea
t
u
r
es
ar
e
ad
d
ed
.
An
o
th
er
p
r
o
b
le
m
p
ar
ticu
lar
l
y
af
f
ec
tin
g
t
h
e
co
m
p
u
tatio
n
ti
m
e
i
s
th
e
lack
i
n
g
s
ca
lab
ili
t
y
o
f
th
e
q
r
n
n
alg
o
r
it
h
m
.
Sev
er
al
ap
p
r
o
ac
h
es
to
th
e
v
ar
i
ab
le
s
elec
tio
n
p
r
o
b
le
m
u
s
i
n
g
i
n
f
o
r
m
atio
n
t
h
eo
r
etic
cr
iter
ia
h
av
e
b
ee
n
p
r
o
p
o
s
ed
.
Ma
n
y
r
el
y
o
n
e
m
p
ir
ica
l e
s
ti
m
a
tes o
f
th
e
m
u
t
u
al
in
f
o
r
m
a
tio
n
b
et
w
ee
n
ea
ch
v
ar
iab
le
an
d
th
e
tar
g
et:
(
∑
(
∑
)
)
(
4
)
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8938
IJ
-
AI
Vo
l.
3
,
No
.
4
,
Dec
em
b
er
20
1
4
:
1
5
0
–
1
5
5
154
∫
∫
(
6
)
W
e
a
r
e
s
e
lec
te
d
th
e
f
i
r
s
t
6
in
f
o
r
m
ativ
e
c
o
m
p
o
u
n
d
s
.
5.
Q
SA
R
M
O
DE
L
S VA
L
I
DA
T
I
O
N
T
h
e
v
ali
d
ati
o
n
o
f
a
q
s
a
r
r
ela
tio
n
s
h
i
p
is
p
r
o
b
ab
ly
th
e
m
o
s
t
im
p
o
r
tan
t
s
t
e
p
o
f
all
.
T
h
e
v
alid
ati
o
n
esti
m
ates
th
e
r
eli
a
b
ili
ty
an
d
ac
cu
r
ac
y
o
f
p
r
e
d
ic
ti
o
n
s
b
ef
o
r
e
th
e
m
o
d
el
is
p
u
t
in
t
o
p
r
a
cti
c
e.
Po
o
r
p
r
e
d
ict
io
n
s
m
is
g
u
id
e
th
e
d
i
r
e
cti
o
n
o
f
d
r
u
g
d
ev
e
lo
p
m
en
t a
n
d
tu
r
n
d
o
w
n
s
tr
ea
m
ef
f
o
r
ts
m
ea
n
in
g
less
.
T
o
v
er
if
y
m
o
d
el
q
u
a
lity
in
r
eg
r
ess
io
n
task
s
,
p
r
ed
ict
io
n
s
ar
e
m
ad
e
o
n
th
e
t
es
tin
g
s
et
in
o
r
d
e
r
t
o
ch
ec
k
th
e
ag
r
e
em
en
t
b
etw
ee
n
th
e
th
eo
r
e
tic
al
v
alu
es
an
d
ex
p
er
im
en
tal
v
alu
es
b
y
ca
l
cu
latin
g
r
o
o
t
-
m
ea
n
s
q
u
ar
e
er
r
o
r
o
f
p
r
e
d
ic
ti
o
n
(
R
MSE
)
.
√
∑
(
̂
)
(
7
)
W
h
e
r
e
,
̂
,
v
a
lu
es
o
f
th
e
p
r
e
d
i
cte
d
v
alu
es
,
an
d
,
v
alu
es
o
f
th
e
ac
tu
al
v
alu
es
.
Ho
w
ev
er
,
it
i
s
n
ec
ess
a
r
y
to
g
et
a
l
ar
g
e
n
u
m
b
er
o
f
test
in
g
c
o
m
p
o
u
n
d
s
in
o
r
d
e
r
to
d
r
aw
s
tatis
ti
ca
l
ly
co
n
v
in
ci
n
g
co
n
c
lu
s
io
n
.
6.
E
XP
E
RM
E
NT
I
A
L
RE
SUL
T
S
Fig
u
r
e
4
.
QQ
n
o
r
m
p
r
o
b
ab
ilit
y
o
f
d
r
u
g
ac
tiv
i
t
y
.
Fig
u
r
e
5
.
T
h
e
er
r
o
r
cu
r
v
es f
o
r
d
is
tin
ct
q
u
an
t
ile
u
s
in
g
QR
NN
a
n
d
QR
E
G.
I
n
th
i
s
s
ec
t
io
n
w
e
p
r
ese
n
t
t
h
e
i
m
p
le
m
en
ta
tio
n
o
f
QR
N
N
esti
m
ates
o
f
t
h
e
ef
f
ec
t
s
o
f
d
r
u
g
c
o
m
p
o
u
n
d
o
n
ch
a
n
g
e
s
in
d
r
u
g
ac
ti
v
it
y
.
W
e
esti
m
ate
t
h
e
m
o
d
el
b
y
s
tat
is
tical
q
u
a
n
tile
r
e
g
r
ess
io
n
an
d
n
o
n
-
li
n
ea
r
QR
N
N
at
q
u
an
tile
s
0
.
1
th
,
0
.
2
th
,
0
.
3
th
,
0
.
4
th
,
5
0
th
,
0
.
6
th
,
0
.
7
th
,
0
.
8
th
an
d
9
th
.
I
n
th
e
s
o
lu
tio
n
p
r
o
ce
s
s
,
w
e
u
s
ed
“
q
u
a
n
tr
eg
”
an
d
“
q
r
n
n
”
p
ac
k
ag
e
s
in
R
p
r
o
g
r
a
m
m
e
i
m
p
le
m
e
n
ted
lin
ea
r
p
r
o
g
r
a
m
m
in
g
.
T
h
e
QQ
n
o
r
m
al
p
r
o
b
a
b
ilit
y
p
lo
ts
is
s
h
o
w
n
i
n
Fi
g
u
r
e
4
,
w
h
ic
h
is
u
s
ed
to
co
m
p
ar
e
s
a
m
p
le
d
ata
ag
ain
s
t
a
th
eo
r
etica
l
n
o
r
m
al
d
is
tr
ib
u
tio
n
.
Fi
g
u
r
e
5
also
s
h
o
w
s
t
h
at
t
h
e
co
r
r
esp
o
n
d
in
g
er
r
o
r
cu
r
v
es
o
f
t
h
e
QR
NN
an
d
QR
E
G
(
f
o
r
d
if
f
er
e
n
t
)
.
T
h
e
m
in
i
m
u
m
R
MS
er
r
o
r
lo
ca
ted
at
.
Fro
m
th
e
cu
r
v
es,
it
i
s
clea
r
th
a
t
QR
NN
p
er
f
o
r
m
a
n
ce
i
s
m
o
r
e
b
etter
th
an
QR
E
G.
B
y
tr
ain
i
n
g
t
h
e
QR
NN
m
o
d
el
f
o
r
2
0
0
iter
atio
n
w
it
h
n
u
m
b
er
o
f
h
id
d
en
n
o
d
es
eq
u
al
to
5
,
an
d
co
m
p
ar
i
n
g
th
e
ac
ti
v
it
y
o
f
ea
ch
d
r
u
g
s
a
m
p
le
w
i
th
r
ea
l
d
a
ta
(
s
ee
Fi
g
6
.
)
,
th
e
R
M
SE
p
r
ed
ictio
n
er
r
o
r
w
as
d
eter
m
in
ed
b
e
o
n
l
y
0
.
5
9
.
Fr
o
m
th
e
s
e
r
esu
lt
s
it
is
co
n
cl
u
d
ed
th
at
th
e
n
eu
r
al
n
et
w
o
r
k
q
u
an
tile
r
eg
r
es
s
io
n
i
s
s
u
p
er
io
r
f
o
r
a
co
m
p
le
x
n
o
n
li
n
ea
r
p
r
ed
ictio
n
.
Evaluation Warning : The document was created with Spire.PDF for Python.
IJ
-
AI
I
SS
N:
2252
-
8938
Qu
a
n
tile R
eg
r
ess
io
n
N
eu
r
a
l N
etw
o
r
ks B
a
s
ed
P
r
ed
ictio
n
o
f D
r
u
g
A
ctivities
(
Mo
h
a
mme
d
E
.
E
l
-
Telb
a
n
y
)
155
7.
CO
NCLU
SI
O
NS
T
h
e
q
u
an
tile
r
e
g
r
ess
io
n
i
s
b
e
liev
ed
to
o
f
f
er
a
m
o
r
e
co
m
p
l
ete
m
o
d
el
t
h
an
th
e
co
n
v
e
n
tio
n
al
m
ea
n
r
eg
r
ess
io
n
.
Si
n
c
e
its
b
ir
th
,
q
u
an
t
ile
r
eg
r
es
s
io
n
h
a
s
b
ee
n
ap
p
lied
to
in
ter
p
r
et
v
ar
io
u
s
p
r
o
b
lem
s
.
Q
u
a
n
tile
r
eg
r
ess
io
n
is
p
o
p
u
lar
an
d
co
m
p
r
eh
en
s
i
v
e,
b
u
t
it
i
s
r
at
h
er
d
if
f
ic
u
lt
to
ap
p
l
y
it
d
ir
ec
tl
y
f
o
r
f
o
r
ec
asti
n
g
.
Hen
ce
,
th
is
s
tu
d
y
ai
m
s
to
ad
v
an
ce
q
u
an
t
ile
r
eg
r
es
s
io
n
f
o
r
f
o
r
ec
asti
n
g
th
e
d
r
u
g
ac
tiv
ities
u
s
i
n
g
n
o
n
-
li
n
ea
r
n
eu
r
al
n
et
w
o
r
k
m
o
d
el.
T
h
e
r
esu
lt
s
is
en
co
u
r
ag
e
b
u
t
th
er
e
is
n
ee
d
s
t
o
ex
p
lo
r
e
d
if
f
er
en
t
d
ata
s
ets
with
lar
g
e
n
u
m
b
er
o
f
s
a
m
p
les a
n
d
ex
p
lo
r
e
d
if
f
er
en
t
f
ea
t
u
r
e
s
elec
tio
n
m
et
h
o
d
s
.
Fig
u
r
e
6
.
T
h
e
R
MSE
o
f
th
e
t
wo
m
et
h
o
d
at
d
if
f
er
en
t q
u
a
n
tile.
RE
F
E
R
E
NC
E
S
[1
]
.
A
.
L
i,
P
re
c
li
n
ica
l
i
n
v
it
ro
sc
re
e
n
in
g
a
ss
a
y
s
f
o
r
d
ru
g
-
li
k
e
p
r
o
p
e
rti
e
s.
Dr
u
g
Disc
o
v
e
ry
T
o
d
a
y
:
T
e
c
h
n
o
lo
g
ies
,
2
(2
):
1
7
9
-
1
8
5
,
2
0
0
5
.
[2
]
.
C.
Bish
o
p
,
P
a
tt
e
rn
Re
c
o
g
n
it
io
n
a
n
d
M
a
c
h
in
e
L
e
a
rn
in
g
,
S
p
rin
g
e
r,
2
nd
,
2
0
0
6
.
[3
]
.
C.
Bish
o
p
.
Ne
u
ra
l
Ne
tw
o
rk
s f
o
r
P
a
tt
e
rn
Re
c
o
g
n
it
io
n
,
Ox
f
o
rd
,
Ox
f
o
rd
Un
ive
rs
it
y
Pre
ss
.
1
9
9
6
.
[4
]
.
C.
L
ip
in
sk
i,
L
e
a
d
-
a
n
d
d
ru
g
-
li
k
e
c
o
m
p
o
u
n
d
s:
t
h
e
ru
le
-
of
-
f
iv
e
re
v
o
lu
ti
o
n
.
Dru
g
Disc
o
v
e
ry
T
o
d
a
y
:
T
e
c
h
n
o
lo
g
ies
1
(4
):
3
3
7
-
3
4
1
,
2
0
0
4
.
[5
]
.
D.
L
iv
in
g
sto
n
e
.
Da
ta
A
n
a
l
y
sis
f
o
r
Ch
e
m
ists.
-
A
p
p
li
c
a
ti
o
n
s
t
o
QSA
R
a
n
d
Ch
e
m
ica
l
p
ro
d
u
c
t
De
sig
n
.
Ox
fo
rd
Un
ive
rs
it
y
Pre
ss
,
1
9
9
5
.
[6
]
.
D.
Ne
wm
a
n
,
S
.
He
tt
ich
,
C.
Blak
e
,
C.
M
e
rz
.
UCI
Re
p
o
sit
o
ry
o
f
M
a
c
h
in
e
L
e
a
rn
in
g
Da
tab
a
se
s,
D
e
p
t.
In
f
o
rm
a
ti
o
n
a
n
d
Co
m
p
u
ter S
c
ie
n
c
e
,
Un
i
v
.
Ca
l
if
o
rn
ia,
Irv
in
e
,
1
9
9
8
.
[7
]
.
E.
Bu
rk
e
a
n
d
G
.
Ke
n
d
a
ll
,
S
e
a
rc
h
M
e
th
o
d
o
lo
g
ies
In
tro
d
u
c
t
o
ry
Tu
to
rials
in
O
p
ti
m
iza
ti
o
n
a
n
d
De
c
isio
n
S
u
p
p
o
r
t
T
e
c
h
n
iq
u
e
s,
2
nd
(e
d
.
)
S
p
ri
n
g
e
r,
2
0
1
4
.
[8
]
.
Ha
n
sc
h
,
A
Qu
a
n
ti
tativ
e
A
p
p
ro
a
c
h
to
Bi
o
c
h
e
m
ica
l
S
t
ru
c
tu
re
-
A
c
ti
v
it
y
Re
latio
n
sh
ip
s.
Acc
t
.
C
h
e
m.
R
e
s.
2
:
2
3
2
-
2
3
9
,
1
9
6
9
.
[9
]
.
J.
De
v
il
lers
.
Ne
u
ra
l
Ne
tw
o
rk
s an
d
Dru
g
De
sig
n
.
A
c
a
d
e
m
i
c
P
re
ss
,
1
9
9
9
.
[1
0
]
.
J.
Ge
rtru
d
e
sa
,
V
.
M
a
lt
a
ro
l
lo
b
,
R.
S
il
v
a
a
,
P
.
Oliv
e
iraa
,
K.
Ho
n
ó
ri
o
a
a
n
d
A
.
d
a
S
il
v
a
.
M
a
c
h
in
e
L
e
a
rn
in
g
T
e
c
h
n
iq
u
e
s
a
n
d
Dru
g
De
sig
n
,
C
u
rr
e
n
t
M
e
d
ici
n
a
l
C
h
e
mistry
,
1
9
,
4
2
8
9
-
4
2
9
7
,
2
0
1
2
.
[1
1
]
.
J.
Hirst,
R.
Kin
g
a
n
d
M
.
S
ter
n
b
e
rg
.
Qu
a
n
ti
tativ
e
S
tru
c
tu
re
-
A
c
ti
v
it
y
Re
latio
n
sh
ip
s
By
Ne
u
ra
l
Ne
tw
o
rk
s
a
n
d
In
d
u
c
ti
v
e
L
o
g
ic P
ro
g
ra
m
m
in
g
.
I.
T
h
e
In
h
ib
it
i
o
n
Of
Dih
y
d
ro
f
o
late
Re
d
u
c
tas
e
b
y
P
y
ri
m
id
in
e
s
.
J
o
u
rn
a
l
o
f
C
o
mp
u
ter
-
Ai
d
e
d
M
o
lec
u
la
r De
sig
n
,
v
o
l
.
8
,
n
o
.
4
,
p
p
.
4
0
5
-
4
2
0
,
1
9
9
4
.
[1
2
]
.
J.
T
a
y
lo
r.
Qu
a
n
ti
le
Re
g
re
ss
io
n
N
e
u
ra
l
Ne
tw
o
rk
A
p
p
ro
a
c
h
t
o
Est
i
m
a
ti
n
g
T
h
e
Co
n
d
it
io
n
a
l
De
n
sity
o
f
M
u
lt
i
p
e
rio
d
Re
tu
rn
s
.
J
o
u
rn
a
l
o
f
F
o
re
c
a
stin
g
,
V
o
l
.
1
9
,
p
p
.
2
9
9
-
3
1
1
,
2
0
0
0
.
[1
3
]
.
L
.
Ch
in
Ye
e
a
n
d
Y.
Ch
u
n
W
e
i,
Cu
rre
n
t
M
o
d
e
li
n
g
M
e
th
o
d
s
Us
e
d
in
QSA
R/QS
P
R,
in
S
tatisti
c
a
l
M
o
d
e
ll
i
n
g
o
f
M
o
lec
u
lar
De
sc
rip
to
rs
in
QSA
R
/QS
P
R,
1
st,E
d
it
e
d
b
y
M
.
De
h
m
e
r,
K.
V
a
rm
u
z
a
,
a
n
d
D.
Bo
n
c
h
e
v
,
W
il
e
y
-
V
CH
V
e
rlag
Gm
b
H &
Co
,
2
0
1
2
.
[1
4
]
.
M
.
El
-
T
e
lb
a
n
y
.
T
h
e
P
re
d
ictiv
e
Lea
rn
in
g
Ro
le
in
Dru
g
De
sig
n
.
In
Jo
u
rn
a
l
o
f
Em
e
r
g
in
g
T
r
e
n
d
s
in
Co
m
p
u
ti
n
g
a
n
d
In
f
o
rm
a
ti
o
n
S
c
ien
c
e
s,
J
ET
CI
S
,
Vo
l.
5
,
No
.
3
M
a
rc
h
2
0
1
4
.
[1
5
]
.
M
.
M
o
h
ri
,
A
.
Ro
sta
m
iza
d
e
h
,
a
n
d
A
.
T
a
l
w
a
l
k
a
r,
F
o
u
n
d
a
ti
o
n
s o
f
M
a
c
h
in
e
L
e
a
rn
in
g
,
M
IT
P
re
ss
,
2
0
1
2
.
[1
6
]
.
P
.
L
e
e
so
n
,
A
.
Da
v
is,
J.
S
tee
le,
Dru
g
-
li
k
e
p
ro
p
e
rti
e
s:
g
u
i
d
in
g
p
r
i
n
c
ip
les
f
o
r
d
e
sig
n
–
o
r
c
h
e
m
ica
l
p
re
ju
d
ice
?
Dru
g
Disc
o
v
e
r
y
T
o
d
a
y
:
T
e
c
h
n
o
l
o
g
ies
,
1
(3
):
1
8
9
-
1
9
5
,
2
0
0
4
.
[1
7
]
.
R.
Bu
rb
i
d
g
e
,
M
.
T
ro
tt
e
r,
B.
Bu
x
t
o
n
a
n
d
S
.
Ho
ld
e
n
.
Dru
g
De
sig
n
b
y
M
a
c
h
in
e
Lea
rn
in
g
:
S
u
p
p
o
rt
Ve
c
to
r
M
a
c
h
in
e
s
f
o
r
P
h
a
rm
a
c
e
u
ti
c
a
l
Da
ta
A
n
a
l
y
sis
.
Co
mp
u
ter
s a
n
d
C
h
e
mistry
,
v
o
l.
2
6
,
n
o
.
1
,
p
p
.
4
-
1
5
,
2
0
0
1
.
[1
8
]
.
R.
Ko
e
n
k
e
r
a
n
d
J.
G
il
b
e
rt
Ba
ss
e
tt
,
Re
g
re
ss
io
n
q
u
a
n
ti
les
,
Ec
o
n
o
me
trica
,
4
6
,
p
p
.
3
3
–
5
0
.
,
1
9
7
8
.
[1
9
]
.
R.
Ko
e
n
k
e
r,
Qu
a
n
t
il
e
Re
g
re
ss
io
n
,
Ca
mb
ri
d
g
e
Un
ive
rs
it
y
Pre
ss
,
Ca
m
b
rid
g
e
,
2
0
0
5
.
[2
0
]
.
S
.
M
a
rsla
n
d
M
a
c
h
i
n
e
L
e
a
rn
in
g
An
A
lg
o
rit
h
m
ic
P
e
rsp
e
c
ti
v
e
,
(
Ch
a
p
m
a
n
&
Ha
ll
/CRC,
2
0
0
9
.
[2
1
]
.
W
.
Du
c
h
,
K.
S
w
a
m
in
a
th
a
n
a
n
d
J.
M
e
ll
e
r,
A
rti
f
icia
l
In
telli
g
e
n
c
e
A
p
p
ro
a
c
h
e
s
f
o
r
Ra
ti
o
n
a
l
Dru
g
De
sig
n
a
n
d
Disc
o
v
e
r
y
,
Cu
rr
e
n
t
Ph
a
rm
a
c
e
u
ti
c
a
l
De
sig
n
,
1
3
,
2
0
0
7
.
Evaluation Warning : The document was created with Spire.PDF for Python.