Laz
Learn
n
or Pre
ct
ve Tox
co
o
atom
tr
,
,o,-
. atom
tr
,
,n,
. atom
tr
,
,o,
. atom
tr
,
,c,
.
atom(tr339,
,c,0). atom(tr339,6,c,0). atom(tr339,7,c,0). atom(tr339,8,o,0).
atom(tr339,9,c,0). atom(tr339,10,n,0). atom(tr339,11,c,0). atom(tr339,12,h,0).
atom(tr339,13,h,0). atom(tr339,14,h,0). atom(tr339,15,h,0). atom(tr339,16,h,0).
atom(tr339,17,h,0). bond(tr339,1,2,1). bond(tr339,2,3,2). bond(tr339,2,4,1).
ond(tr339,4,5,1). bond(tr339,5,6,2). bond(tr339,5,12,1). bond(tr339,6,7,1).
ond(tr339,6,13,1). bond(tr339,7,8,1). bond(tr339,7,9,2). bond(tr339,8,14,1).
ond(tr339,9,10,1). bond(tr339,9,11,1). bond(tr339,10,15,1). bond(tr339,10,16,1).
ond(tr339,11,17,1)
tomcoord(tr339,1,3.0918,-0.8584,0.0066). atomcoord(tr339,2,2.3373,0.0978,0.006).
tomcoord(tr339,3,2.7882,1.2292,0.0072). atomcoord(tr339,4,0.8727,-0.1152,-0.0023).
tomcoord(tr339,5,0.3628,-1.4003,-0.0094). atomcoord(tr339,6,-1.0047,-1.6055,-0.0172).
tomcoord(tr339,7,-1.868,-0.5224,-0.0174). atomcoord(tr339,8,-3.2132,-0.7228,-0.0246).
tomcoord
tr339,9,-1.355,0.7729,-0.0098
. atomcoord
tr339,10,-2.2226,1.8712,-0.0096
.
tomcoord
tr339,11,0.018,0.971,0.0028
. atomcoord
tr339,12,1.0343,-2.2462,-0.0092
tomcoord
tr339,13,-1.3998,-2.6107,-0.0234
. atomcoord
tr339,14,-3.4941,-0.7673,0.8996
.
tomcoord
tr339,15,-3.1824,1.7311,-0.0147
. atomcoord
tr339,16,-1.864,2.7725,-0.0043
.
tomcoord
tr339,17,0.419,1.9738,0.0087
Fig. 1. Representation of the chemical compound TR-339 using Horn clauses
o
a com
oun
re
ationa
representation
.Int
e
o
ow sect
ons we
r
e
ex
a
nt
ese re
resentat
ons (
eta
s can
e
oun
at
ttp://www.informatik.
uni-freiburg.de/˜ml/ptc/ an
t
en we w
ntro
uce our own re
resenta-
ion based on the chemical ontolo
used b
the experts
AR and Qualitative SAR (QSAR) use e
uation sets that allow the
redic
ion of some
ro
erties of the molecules before the ex
erimentation in the labo
ator
. In anal
tical chemistr
, these equations are widel
used to predict spectro
copic, chromato
raphic and some other properties of chemical compounds. Ther
is a number of commercial tools allowing the generation of these descriptors
CODESSA [22], TSAR (Oxford molecular products,
tt
://www.accelr
s.
com/chem/
, DRAGON htt
://www.disat.inimib.it/chm/Dra
on.
htm), etc. These tools represent a chemical compound as a set of attribute valu
airs. This kind of representation is calle
ro
ositiona
in ML. For instance
the de-
cription of a car using propositional description is the following
size, medium
builder, BMW
,
model, 2
0
,
color, white
In addition to the knowledge about a particular compound, it is also useful t
handle general chemical knowledge, what is calle
background knowledg
in ML.
Automatic methods that use background knowledge often consider compounds as a
tructure compose
o
su
structures. T
s
n
o
representat
on
sca
e
e
ationa
ecause an o
ect
s represente
yt
ere
at
ons
ps
etween t
e
r component e
e
ments. For
nstance, a car can
e
escr
e
compose
o
su
parts
et
ec
ass
san
een
ne. In turn, eac
one o
t
ese parts can
e
escr
e
t
e
r own su
compo
nen
s.
A
orm o
re
at
ona
re
resentat
on
ogic programming
t
at re
resents t
e re-
at
ons amon
e
ements
a set o
pre
cates. T
us, a set o
pre
cates can
e use
o esta
s
t
ere
at
ons
p amon
t
e atoms o
amo
ecu
ean
a
so
an
e
as
n
ormat
on a
out t
e compoun
s (suc
as mo
ecu
ar we
t, e
ectr
ca
c
ar
e, etc.)
.1s
ows t
e representat
on o
t
ec
em
ca
compoun
TR-339 (t
-
m
n
-
itrop
eno
o
t
e NTP
ata set. In t
sre
resentat
on, t
ere are t
ree
re
cates