DESIRABILITY OF POPs ACCORDING TO THEIR

1
QSAR MODELLING OF PERSISTENT ORGANIC POLLUTANTS
MOBILITY
2r/P004
PAOLA GRAMATICA and STEFANO POZZI QSAR Research
Unit, Dep. of Structural and Functional Biology,
University of Insubria, via Dunant 3, I - 21100,
Varese (Italy)
e-mail
gramati_at_imiucca.csi.unimi.it http//andromeda.var
bio.unimi.it/QSAR
2
1
MOLECULAR DESCRIPTORS For a large number of
POPs there are great gaps in the experimental
data of several physico-chemical properties, thus
there is a need to develop statistical models to
predict such physical-chemical properties
(boiling point, melting point, logKow, logKoc,
Henrys law constant, TSA, Vmol, water
solubility, vapour pressure) and the atmospheric
half-life for these compounds this can be done
by the QSAR/QSPR approach and the structural
representation of many compounds has been
realised using different kinds of molecular
descriptors. Molecular descriptors represent the
way chemical information contained in the
molecular structure is transformed and coded, in
order to deal with chemical, pharmacological and
toxicological problems in QSAR and QSPR studies.
Molecular descriptors take different aspects of
the chemical information into account. Among the
theoretical descriptors the best known are
molecular weight and count descriptors
(1D-descriptors, i. e. counting of bonds, atoms
of different kind, presence or counting of
functional groups and fragments, etc.). There are
obtained from knowledge of the formula whereas
graph-invariant descriptors (2D-descriptors
including both topological and information
indices), obtained from the knowledge of the
molecular topology. New molecular descriptors
(WHIM 1) contain information about the whole
3D-molecular structure in terms of size, symmetry
and atom distribution. These indices are
calculated 2 from the (x,y,z)-coordinates of a
three-dimensional structure of a molecule,
usually from a spatial conformation of minimum
energy 37 non-directional (or global) and 66
directional descriptors are obtained. Our
representation of a chemical is based on a lot of
molecular descriptors, thus an effective variable
selection strategy is necessary. GA-VSS (Genetic
Algorithm - Variable Subset Selection) was
applied to the whole set of descriptors in order
to set out the most relevant variables in
modelling the POPs chemical-physical properties
and atmospheric half-life by Ordinary Least
Squares regression (OLS) 3. For all the
physical-chemical properties models with good
prediction power are obtained (see poster
Modelling of physico-chemical properties for
organic pollutants- 2r/P002- for further
information) 1 R.Todeschini and P.Gramatica,
3D-modelling and prediction by WHIM descriptors.
Part 5. Theory development and chemical meaning
of the WHIM descriptors, Quant.Struct.-Act.Relat.,
16 (1997) 113-119. 2 R. Todeschini, WHIM-3D /
QSAR - Software for the calculation of the WHIM
descriptors. rel. 4.1 for Windows, Talete srl,
Milan (Italy) 1996. Download http//www.disat.un
imi.it/chm. 3 R. Todeschini, Moby Digs -
Software for Variable Subset Selection by Genetic
Algorithms. Rel. 1.0 for Windows, Talete srl,
Milan (Italy) 1997.
INTRODUCTION An understanding of how persistent
organic pollutants (POPs), particularly PAH, PCB,
dibenzodioxines and pesticides, migrate through
the global environment has become a principal
target to predict their environmental fate, and
consequently their environmental risk. A common
characteristic of most POPs is that they break
down very slowly. Their persistence in the
environment and the fat-solubility of most POPs
allows them to pass along the food chain and
accumulate in animals. These chemicals have toxic
effects on animal reproduction, development and
immunological function, and some are also
probably carcinogenic. The use of POPs in
developed countries has been restricted or even
banned, largely due to the knowledge that these
compounds can move thousands of Kilometres from
the point of release. The most probable way of
migration is the one known as the grasshopper
effect, that consists in migration to higher
latitudes in a series of relatively short jumps
(Figure 1). POPs of different volatility
migrate through the global atmosphere at
different velocities. An individual compound may
have its own distinctive environmental
condensation temperature range, and consequently
its own latitudinal range of condensation. These
compounds are hazardous, due to both their
toxicity for different organisms and their
physical-chemical properties that determine their
environmental fate, mainly persistence,
bioaccumulation, etc.
GRASSHOPPER EFFECT
High mobility
Relat. high mobility
HCB
DDT
Relat. low mobility
Low mobility
3
PRINCIPAL COMPONENT ANALYSIS The biplot of
principal component analysis (Figure 2) for 87
POP (Tab. 1), described by the principal
physico-chemical properties (boiling point,
melting point, logKow, logKoc, Henrys law
constant, TSA, Vmol, water solubility, vapour
pressure) and the atmospheric half-life, shows a
particular distribution of compounds along the
first component (PC1, EV 70.4) according to
their own class-mobility assigned by Mackay and
Wania 4. Consequently, it is possible to use
PC1 to classify all 87 POPs in one of four
classes of mobility (high, relatively high,
relatively low and low mobility). This model does
not consider the atmospheric half-life, because
this property is represented only in the second
component. 4 Frank Wania and Donald Mackay,
Environmental Science Technology, Vol. 30, NO.
9, 1996
FIGURE 1

• DESIRABILITY OF POPs ACCORDING TO THEIR
• ATMOSPHERIC MOBILITY
• The main goal pursued in this work is the
  formulation of a POP ranking by atmospheric
  mobility. Some factors must be considered because
  they influence the atmospheric mobility, in
  particular the compounds half-life and their
  sorption in the atmospheric particle. A
  chemometric strategy known as Multicriteria
  Decision Making, in particular a linear
  desirability functions, was used for this
  purpose.
• POPs with low mobility are considered the most
  desirable.
• The used criteria are
• the first principal component (PC1) score values
  as mobility indicator (optimum low values)
• the logKoc as atmospheric particle sorption
  indicator (optimum high values)
• the half-life values (optimum low values)
• The desirability values for each compounds are
  reported in Tab. 1.
• In figure 3 the desirability values were plotted
  on a multidimensional scaling for the first six
  principal components (Cum Ev 86.3) obtained by
  a principal component analysis for 87 POPs
  described from 162 molecular descriptors.
• The compounds mobility trend seems to be similar
  to their real world distribution.

4
Classification from Mackay and Wania
Chemical classes
Some variables that are represented in the first
principal component
Low mobility
FIGURE 2
5
CLASSIFICATION POP classification according to
mobility was made by three different
classification methods (Classification And
Regression Tree, CART, K-Nearest Neighbours,
K-NN, and Regular Discriminant Analysis, RDA).
The prior classes were obtained from the
desirability values high mobility (1) ? 0.33,
relatively high (2) 0.330, 0.5, relatively low
(3) 0.5, 0.67, low mobility (4) gt 0.67. All the
Classification methods give models with
satisfactory prediction power. The simplest
model, and consequently the most directly
applicable, is the one developed with CART
(figure 4) the selected descriptors are mainly
related to molecular size. Tab. 2 shows the prior
classes (CLA) and the predicted classes with each
model for all POPs. It can be noted that most of
the compounds have been assigned to the same
class by all the applied classification methods,
only compounds at the border of two contiguous
classes have a different classification,
nevertheless, no one compound has been assigned
to not-adjacent classes.
High mobility
Low mobility
Desirability values high values low mobility
FIGURE 3
RDA
? 0.5 ? 0.0 MRcv 14.94
SWLDA
MOLECULAR DESCRIPTORS CHI0 IAC DELS ROUV MAXDP
MW NCl L1m E3u P1v L2s Ts Tm Vu Av
CLASSIFICATION NOMMR 63.22
K 3 MRcv 13.79
KNN
CART (162 DESCRIPTORS)
MRcv 12.64
Selected molecular
descriptors C count descriptors T
topological descriptors W-DIR directional WHIM
descriptors W-ND no directional WHIM
descriptors MW molecular weight NAT
number of atoms (C) NBO number of bonds (C)
NCl number of Cl (C) CHI0 connectivity
index of zero-order (T) CHI1A Randic chi-1
(average) (T) GSI Gordon-Scatlebury index
connection number (T) BAL Balaban distance
connectivity index (T) IAC total information
index on atomic composition (T) IDDE mean
information content on the distance degree
equality (T) DELS total electrotopological
difference (T) ROUV Rouvray index (T) MAXDP
maximum positive intrinsic state difference
(T) L1m dimension among the first component
with atomic mass weight (W-DIR) L2s size among
the second component with electrotopological
weight (W-DIR) E1u, E3u density among
respectively the first and the third dimension
with unit weight (W-DIR) P2u shape among the
second component with unit weight (W-DIR) P1v
shape among the first component with van der
Waals volume weight (W-DIR) Ts, Tm size
(eigenvalue sum) with respectively atomic mass
and electrotopological weight (W-ND) Av size
(cross-term eigenvalue sum) with van der Waals
volume weight (W-ND) Vu size (complete
eigenvalue expression) with unit weight (W-ND)
FIGURE 4
TABLE 1
CONCLUSIONS Good QSAR classification models
with satisfactory prediction power
allow molecular descriptors modelling of the
mobility of persistent organic
pollutants. Commonly descriptors are related to
molecular size, so this property seems to be the
most important in POP mobility description.
6
TABLE 1
TABLE 2