This article was downloaded by: [FU Berlin]

On: 07 May 2015, At: 01:11

Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered
office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

SAR and QSAR in Environmental

Research
Publication details, including instructions for authors and
subscription information:
http://www.tandfonline.com/loi/gsar20

Development of ecotoxicity QSAR

models based on partial charge
descriptors for acrylate and related
compounds
a a a
A. Furuhama , Y. Aoki & H. Shiraishi
a
Center for Environmental Risk Research , National Institute for
Environmental Studies (NIES) , Tsukuba , Japan
Published online: 11 Sep 2012.

To cite this article: A. Furuhama , Y. Aoki & H. Shiraishi (2012) Development of ecotoxicity QSAR
models based on partial charge descriptors for acrylate and related compounds, SAR and QSAR in
Environmental Research, 23:7-8, 731-749, DOI: 10.1080/1062936X.2012.719542

To link to this article: http://dx.doi.org/10.1080/1062936X.2012.719542

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the
“Content”) contained in the publications on our platform. However, Taylor & Francis,
our agents, and our licensors make no representations or warranties whatsoever as to
the accuracy, completeness, or suitability for any purpose of the Content. Any opinions
and views expressed in this publication are the opinions and views of the authors,
and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content
should not be relied upon and should be independently verified with primary sources
of information. Taylor and Francis shall not be liable for any losses, actions, claims,
proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or
howsoever caused arising directly or indirectly in connection with, in relation to or arising
out of the use of the Content.

This article may be used for research, teaching, and private study purposes. Any
substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing,
systematic supply, or distribution in any form to anyone is expressly forbidden. Terms &
 Conditions of access and use can be found at http://www.tandfonline.com/page/terms-
and-conditions
Downloaded by [FU Berlin] at 01:11 07 May 2015
 SAR and QSAR in Environmental Research
Vol. 23, Nos. 7–8, October–December 2012, 731–749

Development of ecotoxicity QSAR models based on partial charge

descriptors for acrylate and related compounds
A. Furuhama*, Y. Aoki and H. Shiraishi

Center for Environmental Risk Research, National Institute for Environmental

Studies (NIES), Tsukuba, Japan
(Received 31 January 2012; in final form 11 April 2012)

Using Gasteiger’s partial equalization of orbital electronegativity (PEOE)

method, we constructed ecotoxicity prediction equations based on two-
dimensional descriptors for ,-unsaturated carbonyl compounds. After exam-
Downloaded by [FU Berlin] at 01:11 07 May 2015

ining electrostatic effects on the calculated ecotoxicities of 10 ,-unsaturated

ketones and aldehydes (A-group compounds) by using the Mulliken atomic
charges on the carbonyl oxygen atoms, we investigated the efficacy of the PEOE
descriptors for the same 10 compounds and the correlation between the PEOE
descriptors and the Mulliken charge. We then constructed QSAR models for
acute fish and Daphnia toxicities by using the PEOE descriptors for acrylic acids
and compounds with acrylate-like substructures (CH-group compounds). In the
constructed models, the adjusted squared correlation coefficients between
measured and calculated toxicities with the lowest Akaike information criterion
were 0.77 and 0.79, respectively. The applicability of the constructed models was
then evaluated for various methacrylates and similar compounds (CH3-group
compounds). Both the fish and the Daphnia toxicities of some of the CH3-group
compounds were underestimated by these models. Nevertheless, we concluded
that the QSAR models based on the PEOE descriptors were practical for
predicting acute toxicity, especially for ,-unsaturated carbonyl compounds with
an -hydrogen. Combining hydrophobicity and PEOE descriptors led to accurate
predictions for fish toxicity.
Keywords: ecotoxicity predictions; partial charges; PEOE; ,-unsaturated
carbonyl compounds; acrylate, acute fish and Daphnia toxicities

1. Introduction
Quantitative structure–activity relationship (QSAR) and category-approach models are
practical tools for evaluating the acute aquatic toxicities [1,2] and chemical risks posed by
various environmental pollutants [3–5]. For example, the Organisation for Economic
Co-operation and Development (OECD) QSAR Toolbox [6] permits the categorization of
compounds by various defined rules for predicting ecotoxicity, such as the rules in the US
Environmental Protection Agency’s (US EPA) ECOSAR (Ecological Structure Activity
Relationships) class program [7] and the Verhaar classification rules [8]. The categoriza-
tion of compounds is the first step for predicting toxicity by means of a QSAR or category-
approach model. For example, the Japanese National Institute for Environmental Studies
developed an ecotoxicity QSAR system, referred to as the KAshinhou Tool for Ecotoxicity

*Corresponding author. Email: [email protected]

ISSN 1062–936X print/ISSN 1029–046X online

ß 2012 Taylor & Francis
http://dx.doi.org/10.1080/1062936X.2012.719542
http://www.tandfonline.com
 732 A. Furuhama et al.

Figure 1. Scheme-1wat, which includes an explicit water molecule. R1 ¼ aliphatic or aromatic group;
R2 ¼ H, C5H7, or C2H5; R3 ¼ H or CH3; R4SH ¼ GSH or CH3SH. This figure is a reproduction of
Figure 1 in [20].

(KATE) [9–11], using the Japanese Ministry of the Environment ecotoxicity database [12]
and the US EPA fathead minnow database [13]. In this system, compounds can be entered
according to their one-dimensional structures and be classified by their substructures.
Downloaded by [FU Berlin] at 01:11 07 May 2015

The QSARs in KATE are expressed as simple linear correlations between the log of the
octanol/water partition coefficient (log POW) of a compound and its acute aquatic toxicity
to fish and Daphnia (Daphnia magna) [10].
The descriptors of a QSAR model are defined for each toxic endpoint. In a series of
papers [14–16], Dimitrov and co-workers reported studies on the simulation of the
metabolic pathways of xenobiotic chemicals for regulatory purposes. They introduced new
bioconcentration factor (BCF) baseline models [16], which represented improvements over
their conventional BCF models [17,18], by using pKa or the ratio between distribution and
partition coefficients to estimate the ionization potential of chemicals. Very recently,
Moosus and Maran [19] performed a QSAR analysis with 689 calculated descriptors
including quantum chemical descriptors based on three-dimensional (3D) structures
(e.g. the energy of the highest occupied molecular orbital) for acute toxicity to D. magna.
They concluded that two descriptors (the energy of the highest occupied molecular orbital
and the weighted partial negative surface area) could express the electronic structure of a
chemical and the interaction between the chemical and targets, such as protein and DNA,
in D. magna.
In a previous paper [20], we reported that the toxicity of reactive compounds, such as
,-unsaturated carbonyl compounds (which fall into the Michael addition reaction
domain), cannot be predicted from a hydrophobicity descriptor (e.g. log POW) alone [20]
but can be predicted from reactivity indicators (descriptors), such as the rate constant for
reaction with a biomolecule such as glutathione [21]. The reaction between a compound
and a biomolecule, such as a protein, is often the initial event leading to the toxicity of the
compound and cannot be explained by log POW alone. The Michael addition reaction is
one such reaction that occurs in vitro (in cultured cells or organs) [22–24]. In our previous
paper [20], we reported that the activation energy, defined by density functional theory
calculations, of an initial step of the Michael reaction pathway (designated Scheme-1wat;
see Figure 1, which is a reproduction of Figure 1 in Furuhama et al. [20]) could be used to
assess the acute fish toxicity of acrolein, a Michael acceptor. However, owing to the
complexity of the Michael-type addition and the effects of surrounding amino acid
residues and water molecules, the activation energies obtained with Scheme-1wat were not
practical for predicting fish toxicity with linear regression equations. More intensive and
 SAR and QSAR in Environmental Research 733

Figure 2. Resonance stabilization of the Michael-type addition transition state and solvation-
mediated participation of the -carbon atom in the transition-state (TS) structure (actually taking
place in an aqueous phosphate buffer solution). Dashed lines indicate developing or vanishing bonds
of the resonance structures representing the TS, and dotted lines represent hydrogen bonds between
the TS and solvating H2O and H3Oþ. Reprinted with permission from Chemical Research in
Toxicology 2010, 23, 1576–1585. Copyright 2010 American Chemical Society.
Downloaded by [FU Berlin] at 01:11 07 May 2015

computationally expensive investigations are required to estimate the activation energy of

reaction pathways by quantum chemical calculations for constructing practical QSARs.
In this study, to develop a practical method for predicting the acute toxicities of a series
of ,-unsaturated carbonyl compounds, we concentrated on electrostatic effects,
specifically effects originating from the atomic charge (population) on the carbonyl
oxygen. These effects are based on the influence of resonance stabilization of the transition
state of the Michael-type addition (see Figure 2, which is a reproduction of Figure 4 in
Schwöbel et al. [25]). We also introduced linear regression equations involving Gasteiger’s
partial equalization of orbital electronegativity (PEOE) descriptor [26] without using 3D
molecular structures to include electrostatic charge effects. The PEOE descriptor was
adapted to simple linear QSAR equations and multiregression QSAR equations in which a
PEOE descriptor was combined with log POW. Compounds with methacrylate-like
substructures were used to evaluate the applicability of the constructed QSAR models.

2. Materials and methods

2.1 Datasets
Measured toxicities were obtained from three toxic endpoint datasets: 96-h median lethal
concentrations (LC50) in medaka (Oryzias latipes) obtained from acute toxicity tests; 48-h
median effective concentrations (EC50) in Daphnia (D. magna) obtained from acute
immobilization tests and gathered by the Japanese Ministry of the Environment [12]; and
96-h LC50 data for fathead minnow (Pimephales promelas) obtained by Russom et al. [27]
and accessed from the US EPA fathead minnow database [13] or the OECD QSAR
Toolbox (ver. 2.1) [6]. From the compounds in the 96-h fish (medaka and fathead minnow)
LC50 dataset, we selected 10 ,-unsaturated ketones and aldehydes (designated as the
A-group; Table 1). (Note that the dataset listed in Table 1 is not identical to that used in
our previous work [20]: guppy toxicities were omitted, and data for 10 were added.)
From the same fish LC50 and Daphnia EC50 datasets, we defined two additional groups
of compounds: the CH-group comprised acrylic acids and compounds with acrylate-like
 Downloaded by [FU Berlin] at 01:11 07 May 2015

734

Table 1. Measured fish toxicities, log POW values, and qO and PEOE_PC- descriptors for 10 ,-unsaturated ketones and aldehydes (A-group)a.
A. Furuhama et al.

log (1/LC50
CAS no. Nameb Washed SMILESc [mmol L1]) log POW d qO e PEOE_PC-f

1 107-02-8 Acrolein C(¼O)C¼C 3.52g 0.01 0.378 0.422

2 4170-30-3 Crotonaldehyde C(¼O)C¼CC 2.99h 0.52cp 0.390 0.453
3 623-15-4 4 -(2-Furyl)but-3-en-2-one C(C)(¼O)C¼Cc1ccco1 0.81h 1.35 0.428 0.830
4 122-57-6 Benzalacetone C(C)(¼O)C¼Cc1ccccc1 1.41h 2.18 0.423 0.683
5 20766-36-3 4-(3-Chlorophenyl)-3-buten-2-one CC(¼O)C¼Cc1cccc(Cl)c1 1.78h 2.81 0.418 0.666
6 79-77-6 4-(2,6,6-Trimethyl-1- C1(C)(C)C(C¼CC(C)¼O)¼C(C)CCC1 1.58g 3.72 0.426 0.769
cyclohexen-1-yl)-3-buten-2-one
7 104-55-2 Cinnamaldehyde c1(C¼CC¼O)ccccc1 1.59g,i 1.90 0.396 0.688
8 6203-18-5 4-Dimethyl aminocinnamaldehyde c1(C¼CC¼O)ccc(N(C)C)cc1 1.47h 2.21cp 0.411 0.884
9 122-40-7 Amylcinnamic aldehyde c1(C¼C(C¼O)CCCCC)ccccc1 2.35h 4.47cp 0.398 0.918
10 1335-46-2 Methylionone O¼C(C¼CC1C(C)¼C(CCC1(C)C)C)C 1.84h 4.30 0.422 0.834

Note: aThe acrolein moiety is the key substructure of the A-group (see Figure 3). bChemical names are those used in the Japanese Ministry of the
Environment database [12] when the CAS no. is available in the database. cBased on the SMILESTM strings after explicit hydrogen attachment to the
structure by application of the ‘‘Database Wash’’ functions with the MOE [28] default setting. dValues labelled ‘‘cp’’ are calculated log POW values
(ClogP [40]). eMulliken atomic charge on the carbonyl oxygen atom. fSum of the PEOE negative atomic charges. gValue obtained from Russom et al.
[27]. hValue obtained from the Japanese Ministry of the Environment database [12]. iValue obtained from the OECD QSAR Toolbox [6] ver. 2.1.
 SAR and QSAR in Environmental Research 735

substructures (that is, ,-unsaturated esters with a hydrogen atom on the  carbon;
see Table 2 and Appendix 1 (see supplementary materials, available via the
Supplementary Content tab on the article’s online page at http://dx.doi.org/10.1080/
1062936X.2012.719542); and the CH3-group comprised compounds with methacrylate-like
substructures (that is, ,-unsaturated esters with a methyl group or some other group on
the  carbon; see Table 3 and Appendix 2, available via the Supplementary Content tab on
the article’s online page at http://dx.doi.org/10.1080/1062936X.2012.719542).

2.2 Selection of CH- and CH3-group compounds

We selected the CH-group and CH3-group compounds by using the SMILESTM
(simplified molecular input line entry specification) string C(¼O)C¼C with the string
O(C(¼O)C(C)¼C)C excluded, and the string O(C(¼O)C(C)¼C)C, respectively, and the
Molecular Operating Environment (MOE, ver. 2008.10) software [28]. Other software,
including the OECD QSAR Toolbox, can be used to select compounds with specific
substructures, but we chose to use MOE because it can be used both for compound
Downloaded by [FU Berlin] at 01:11 07 May 2015

selection and for the PEOE calculations, which will be described in section 2.3.
The substituent on the  carbon was the key difference between the CH- and CH3-
group compounds. The CH-group compounds are classified as Acrylates or Acids acrylic
in the current version of KATE (KATE2011) [9], and two of the CH3-group compounds,
28 and 35, are not methacrylates. The CH-group comprised 15 compounds (with fish
toxicity data for all 15 of the compounds and Daphnia toxicity data for 10). The CH3-
group also comprised 15 compounds (with fish toxicity data for 14 and Daphnia toxicity
data for 10). For two of the compounds in the CH3-group, the fish toxicities exceeded
100 mg L1, and for one compound, the Daphnia toxicity exceeded 10 mg L1.
Note that the classification rules for the fish and Daphnia toxicities of Acrylates and
Acids acrylic in KATE2011 are identical to those in the March 2009 version (KATE2009)
[10]. The QSARs for Acids acrylic, which had only three reference compounds in each
class, in KATE2011 were also identical to those of KATE2009. However, the QSARs for
Acrylates were different because the number of reference data was larger. The QSAR for
the fish toxicity of Acrylates in KATE2011 takes the form
 

log 1=LC50 mmol L1 ¼ 0:246ðlog POW Þ þ 1:46 ð1Þ
n ¼ 9 ðn ¼ 6 in KATE2009Þ, RMSE ¼ 0:278, r2 ¼ 0:508,
q2LOO ¼ 0:280, log POW range ½0:210, 3:330,

where n, RMSE, r2, and q2LOO denote the number of compounds classified as Acrylates, the
root mean square error, the squared correlation coefficient, and the leave-one-out version
of the squared correlation coefficient, respectively; log POW range indicates the minimum
and maximum measured log POW values. The QSAR for the Daphnia toxicity of Acrylates
in KATE2011 takes the form
 

log 1=EC50 mmol L1 ¼ 0:122ðlog POW Þ þ 1:34 ð2Þ
n ¼ 5 ðn ¼ 4 in KATE2009Þ, RMSE ¼ 0:175, r2 ¼ 0:343,
q2LOO ¼ 0:034, log POW range ½0:210, 2:620:
 Downloaded by [FU Berlin] at 01:11 07 May 2015

Table 2. Measured fish and Daphnia toxicities, log POW values, and PEOE_PC_# descriptors for 15 compounds with acrylate-like substructures
736

(CH-group)a.

log (1/LC50 log (1/EC50

CAS no. Nameb Washed SMILESc [mmol L1])d [mmol L1])e log POW f PEOE_PC-_#g

11 105-75-9 Dibutyl fumarate O(C(¼O)C¼CC(OCCCC)¼O)CCCC 2.56 N/A 3.96cp 0.72 (2)

12 106-63-8 Isobutyl acrylate O(C(¼O)C¼C)CC(C)C 1.79 N/A 2.22 0.81
13 110-44-1 Sorbic acid O¼C([O-])C¼CC¼CC 0.17 0.20 1.33 1.39
14 13048-33-4 Hexamethylene diacrylate O(C(¼O)C¼C)CCCCCCOC(¼O)C¼C 2.77 1.92 2.62 0.74 (2)
15 140-88-5 Ethyl acrylate O(C(¼O)C¼C)CC 1.92 1.36 1.32 0.71
A. Furuhama et al.

16 141-32-2 n-Butyl acrylate O(C(¼O)C¼C)CCCC 1.73 1.39 2.36 0.81

17 2439-35-2 2 -(Dimethylamino)ethyl O(C(¼O)C¼C)CC[NHþ](C)C 1.23 1.16 0.97cp 1.00
acrylate
18 2499-95-8 Hexyl acrylate O(C(¼O)C¼C)CCCCCC 2.14 N/A 3.33 0.92
19 3066-71-5 Cyclohexyl acrylate O(C(¼O)C¼C)C1CCCCC1 2.02 N/A 2.8 0.87
20 40220-08-4 Tris(2-hydroxyethyl)isocyanuric O¼C1N(CCOC(¼O)C¼C)C 1.80 0.69 2.59cp 1.15 (3)
acid acrylate (¼O)N(CCOC(¼O)C¼C)C
(¼O)N1CCOC(¼O)C¼C
21 5536-61-8 Sodium methacrylate O¼C([O-])C(C)¼C 0.21 0.50 0.93 1.26
22 79-10-7 Acrylic acid O¼C([O-])C¼C 0.07 0.19 0.35 1.24
23 818-61-1 2-Hydroxyethyl acrylate O(C(¼O)C¼C)CCO 1.25 1.35 -0.21 1.06
24 96-33-3 Methyl acrylate O(C(¼O)C¼C)C 1.79 1.52 0.8 0.67
25 999-61-1 2-Hydroxypropyl acrylate O(C(¼O)C¼C)CC(O)C 1.60 N/A 0.35 1.09

Note: aEither the methyl acrylate or acrylic acid moiety is the key substructure of the CH-group (see Appendix 1). bChemical names are those used in
the Japanese Ministry of the Environment database [12]. cBased on the SMILESTM strings after explicit hydrogen attachment to the structure by
application of the ‘‘Database Wash’’ functions with the MOE [28] default setting. All acrylic acids were treated as anions in the PEOE estimation,
whereas neutral SMILESTM strings were used for 13, 17, 21 and 22 to estimate other values, such as ClogP [40]. dAll fish toxicities were obtained from
the Japanese Ministry of the Environment database [12], except the toxicities for 11, 12, 18, 19 and 25, which were obtained from Russom et al. [27].
N/A ¼ data not available. eAll Daphnia toxicities were obtained from the Japanese Ministry of the Environment database [12]. fValues labelled ‘‘cp’’ are
calculated log POW values (ClogP [40]). gPEOE_PC-_# ¼ PEOE_PC- divided by the number of ,-unsaturated carbonyl oxygen atoms in the
compound. The number of unsaturated carbonyl oxygen atoms is indicated in parentheses. For example, 11 and 20 have PEOE_PC- values of 1.44
and 3.46, respectively.
 Downloaded by [FU Berlin] at 01:11 07 May 2015

Table 3. Measured fish and Daphnia toxicities, log POW values and PEOE_PC_# descriptors for 15 compounds with methacrylate-like substructures
(CH3-group)a.

log (1/LC50 log (1/EC50

CAS no. Nameb Washed SMILESc [mmol L1])d [mmol L1])e log POW f PEOE_PC-_#g

26 105-16-8 2 -(Diethylamino)ethyl methacrylate O(C(¼O)C(C)¼C)CC[NHþ](CC)CC 4100h 0.29 2.33cp 1.06

27 106-91-2 Glycidyl methacrylate O1CC1COC(¼O)C(C)¼C 1.71 0.75 0.74cp 1.06
28 11070-44-3 Tetrahydromethylphthalic anhydride O1C(¼O)C¼2 C(CCCC¼2C)C1¼O 4100h 0.11 1.30cp 0.99
29 2370-63-0 2-Ethoxyethyl methacrylate O(C(¼O)C(C)¼C)CCOCC 0.76 N/A 1.36 1.11
30 2455-24-5 Tetrahydrofurfuryl methacrylate O1CCCC1COC(¼O)C(C)¼C 0.69 N/A 1.38 1.12
31 2495-37-6 Benzyl methacrylate O(Cc1ccccc1)C(¼O)C(C)¼C 1.58 N/A 2.50 1.01
32 2867-47-2 2 -(Dimethylamino)ethyl methacrylate O(C(¼O)C(C)¼C)CC[NHþ](C)C 0.92 0.68 1.28cp 1.03
33 352-87-4 2,2,2-Trifluoroethyl methacrylate FC(F)(F)COC(¼O)C(C)¼C 1.31 0.88 1.56 1.19
34 4655-34-9 Isopropyl methacrylate O(C(C)C)C(¼O)C(C)¼C 0.53 N/A 2.25 0.77
35 50512-35-1 Diisopropyl S1CCSC1¼C(C(OC(C)C)¼O) 1.49 410h 3.32 0.76 (2)
1,3-dithiolan-2-ylidenemalonate C(OC(C)C)¼O
36 688-84-6 2-Ethylhexyl methacrylate O(C(¼O)C(C)¼C)CC(CCCC)CC 1.85 1.63 4.24 1.05
37 80-62-6 Methyl methacrylate O(C(¼O)C(C)¼C)C N/A 0.08 1.38 0.70
38 868-77-9 2-Hydroxyethyl methacrylate O(C(¼O)C(C)¼C)CCO 0.24 0.47 0.47 1.07
39 96-05-9 Allyl methacrylate O(C(¼O)C(C)¼C)CC¼C 2.11 N/A 1.48 0.86
40 97-88-1 n-Butyl methacrylate O(C(¼O)C(C)¼C)CCCC 1.40 0.76 2.88 0.84

Note: aThe methyl methacrylate moiety is the key substructure of the CH3-group (see Appendix 2). bChemical names are those used in the Japanese
Ministry of the Environment database [12]. cBased on the SMILESTM strings after explicit hydrogen attachment to the structure by application of the
‘‘Database Wash’’ functions with the MOE [28] default setting. The compounds were treated as ions in the PEOE estimation, whereas neutral
SMILESTM strings were used for 26 and 32 to estimate other values, such as ClogP [40]. dAll fish toxicities were obtained from the Japanese Ministry of
the Environment database [12], except the toxicities for 29, 30, 31, 34, 38 and 39, which were obtained from Russom et al. [27]. N/A ¼ data not
available. Because the LC50 value for 38 from the Japanese Ministry of the Environment database [12] was 4 100 mg L1, the value of Russom et al.
[27] was used. eAll Daphnia toxicities were obtained from the Japanese Ministry of the Environment database [12]. fValues labelled ‘‘cp’’ are calculated
log POW values (ClogP [40]). gPEOE_PC-_# ¼ PEOE_PC- divided by the number of ,-unsaturated carbonyl oxygen atoms in the compound. The
number of unsaturated carbonyl oxygen atoms is indicated in parentheses. hThe LC50 and EC50 values from the Japanese Ministry of the Environment
database were LC50 4 100 mg L1 and EC50 4 10 mg L1, respectively.
SAR and QSAR in Environmental Research
737
 738 A. Furuhama et al.

Figure 3. Energy minimum structures of the A-group compounds (1–10) used to calculate the
Mulliken atomic charges.
Downloaded by [FU Berlin] at 01:11 07 May 2015

2.3 Descriptor calculations

The Mulliken atomic charges on the carbonyl oxygens (qO) of the A-group compounds
were calculated after geometrical optimization at the DFT B3LYP/6-31 G** level of
theory with Gaussian 09 A02 [29]. The initial 3D structures for the quantum chemical
calculations were based on structures obtained with the CORINA software program
(an automatic 2D-to-3D conversion program) [30,31] from SMILESTM strings [32] for the
DFT-optimized geometries (Figure 3).
The PEOE partial charges and the descriptor PEOE_PC- (which is the sum of the
negative charges in a compound) [33] of the A-, CH-, and CH3-group compounds were
calculated with the MOE software [28] after explicit hydrogen attachment on the
structures from the SMILESTM strings by application of the ‘Database Wash’ functions
with the MOE default setting. Unlike other calculated atomic charges, the PEOE partial
charges are defined without using 3D molecular structures. Some charge values based on
quantum chemical calculations depend on the level of theory and the basis set [34], for
example, Mulliken atomic charge analysis. The PEOE method is described in Gasteiger
[35], which also reviews results obtained by means of the PEOE method, such as the
correlation between PEOE and experimental properties such as C1s-ESCA shifts and
dipole moments [35]. In addition, under the assumption that the number of the carbonyl
oxygens is related to reactivity and toxicity, we introduced the descriptor PEOE_PC-_#,
which is PEOE_PC- divided by the number of ,-unsaturated carbonyl oxygen atoms in a
compound. PEOE_PC- was identical to PEOE_PC-_# for 1–10 (Table 1) because these
compounds contain one ,-unsaturated carbonyl oxygen. Tables 2 and 3 show
PEOE_PC-_# values and numbers of carbonyl oxygen atoms in parentheses.

2.4 Evaluation of the PEOE-based equations

We evaluated the equations based on multiregression analyses within the same reference
dataset used for the simple linear regression analyses, by using the Akaike information
criterion (AIC) calculated by the R software [36].
 SAR and QSAR in Environmental Research 739
Downloaded by [FU Berlin] at 01:11 07 May 2015

Figure 4. Correlations between the fish toxicity of the A-group compounds and qO (Mulliken atomic
charge on carbonyl oxygen) (a) and the PEOE_PC- descriptor (b).

Recently reported methodology for electronegativity [37,38] and the Spectral-SAR

algorithm [39] might also be useful for ecotoxicity predictions. The PEOE method,
adopted in this study, is often used for constructing descriptors for ecotoxicity predictions.

3. Results and discussion

3.1 Fish toxicity and charge descriptors for the A-group compounds
We plotted fish toxicity, as indicated by log (1/LC50 [mmol L1]), against qO or PEOE_PC-
for the A-group compounds (Figure 4) and obtained linear regression equations for the
data (Table 4). Fish toxicity tended to increase as the magnitude of the charge descriptor
increased. When we defined 7, which is an aldehyde with an aromatic ring on the  carbon,
as an outlier in the qO data (n ¼ 9), the squared correlation coefficients (r2 and r2adj ) and the
RMSE [standard error, SE] were higher and lower than the corresponding values for
n ¼ 10, respectively. That q2LOO increased from 0.57 to 0.83 also indicated the robustness of
the equation after omission of 7. In addition, when we defined 7–9, all of which are
aldehydes with an aromatic ring on the  carbon, as outliers in the qO data (n ¼ 7), r2
(as well as r2adj and q2LOO ) and RMSE [SE] for n ¼ 7 were also higher and lower,
respectively, than for n ¼ 10. This result indicates that the robustness of the linear equation
based on the qO descriptor was enhanced by omission of aldehydes with an aromatic group
on the  carbon. The aromatic ring may have increased the charge on the carbonyl oxygen
 740 A. Furuhama et al.

Table 4. Simple linear regression equations for correlation between fish toxicity and Mulliken
charge or PEOE descriptor for A-group compounds (1–10).

RMSE RMSE_LOO
n r2 (r2adj ) [SE] q2LOO [SELOO] Outliers

log(1/LC50) ¼ 39.85qO þ 18.23 10 0.73 (0.70) 0.39 [0.44] 0.57 0.50 [0.56]
log(1/LC50) ¼ 45.00qO þ 20.44 9 0.89 (0.87) 0.27 [0.31] 0.83 0.33 [0.38] 7
log(1/LC50) ¼ 46.42qO þ 21.12 7 0.94 (0.92) 0.22 [0.26] 0.89 0.29 [0.34] 7, 8, 9
log(1/LC50) ¼ 3.33(PEOE_PC-) þ 4.32 10 0.49 (0.43) 0.54 [0.61] 0.18 0.72 [0.81]
log(1/LC50) ¼ 4.49(PEOE_PC-) þ 4.99 9 0.76 (0.73) 0.39 [0.44] 0.59 0.51 [0.58] 9
log(1/LC50) ¼ 5.01(PEOE_PC-) þ 5.33 7 0.81 (0.77) 0.38 [0.45] 0.63 0.54 [0.64] 7, 8, 9
log(1/LC50) ¼ 4.98(PEOE_PC-) þ 5.25 8 0.84 (0.81) 0.34 [0.39] 0.67 0.49 [0.57] 9, 10

Notes: The unit for LC50 is mmol L1. qO and PEOE_PC- denote the Mulliken atomic charge on the
carbonyl oxygen and the sum of the PEOE negative atomic charges, respectively. n, r2 (r2adj ), RMSE
[SE], q2LOO , RMSE_LOO [SELOO], and Outlier denote number of chemicals, squared correlation
coefficient (with adjustment for the degrees of freedom), root mean square error [standard error],
leave-one-out version of the squared correlation coefficient, leave-one-out version of RMSE [SE],
and outliers for the regression equations, respectively.
Downloaded by [FU Berlin] at 01:11 07 May 2015

of the aldehyde compounds. Nevertheless, because the equation for qO for the 10
,-unsaturated ketones and aldehydes had a sufficiently high r2 (0.7) and a small RMSE
(50.5), the charge on the carbonyl oxygen was more reliable for predicting fish toxicity
than activation energy or log POW [20]. Therefore, we included multiregression equations
that combine log POW and the charge descriptors only in the supplemental information
(Appendix 3, available via the Supplementary Content tab on the article’s online page at
http://dx.doi.org/10.1080/1062936X.2012.719542). Moreover, because the main purpose
of this article is to introduce a descriptor without using 3D structures (which are necessary
for performing quantum chemical calculations), we do not discuss other atomic charge
population analyses based on quantum chemical calculations, such as natural atomic
orbital or electrostatic potential calculations [34]. The diversities of 3D structure
conformations and high-level quantum chemical calculations present some problems
related to conformer selection and are time consuming, respectively, and therefore are not
easy to use for a toxicity prediction system.
The correlation between fish toxicity and PEOE_PC- did not show the same behaviour
as the correlation between fish toxicity and qO (Figure 4b): for example, the presence of an
aromatic ring may not increase the charge on the carbonyl oxygen of aldehydes.
The PEOE_PC- (¼ the sum of negative charges in a compound) descriptors for 7 and 9
were more negative than the corresponding qO descriptors, but the negative shift was
greater for 9, which has a pentyl group on the  carbon, than for 7 (Figure 4), because the
number of negatively charged atoms in 9 is higher than the number in 7. The PEOE partial
charges on the atoms of 7 (Figure 5) were almost identical to the charges on the
corresponding atoms of 9, except for the  carbon and the atoms connecting to it. In other
words, PEOE_PC- depended on the number of carbons in the  side chain. Thus, unlike
the QSAR equation based on the qO descriptor, the QSAR equation based on the
PEOE_PC- descriptor was strongly affected by the nature of the substructure on the 
carbon of A-group ,-unsaturated compounds (see Appendix 3 for the statistical analysis
based on AIC). The correlations between fish toxicity and qO and PEOE_PC- (Figure 6
and Appendix 4, available via the Supplementary Content tab on the article’s online page
 SAR and QSAR in Environmental Research 741
Downloaded by [FU Berlin] at 01:11 07 May 2015

Figure 5. PEOE partial charges of 7 and 9 calculated by MOE software [28].

Figure 6. Correlation between qO (Mulliken atomic charge on the carbonyl oxygen) and PEOE_PC-
for the A-group compounds.
 742 A. Furuhama et al.

at http://dx.doi.org/10.1080/1062936X.2012.719542) indicated that the negative shift of

PEOE_PC- was highest for 9 and that the shifts for 7 and 8 were higher than those for the
other seven compounds.
On the basis of the results described in this section, we concluded that the two charge-
based descriptors correlated with acute fish toxicity but showed different behaviour with
regard to the outlier compounds with specific substructures.

3.2 Construction of QSAR models for the CH-group compounds

The QSAR models including the PEOE descriptor and log POW (measured or calculated
ClogP [40]) for the fish and Daphnia toxicities of the CH-group compounds, expressed as
log(1/LC50 [mmol L1]) and log(1/EC50 [mmol L1]), respectively, are summarized in
Table 5. As was the case for the A-group compounds, both the fish and the Daphnia
toxicities tended to increase as the charge descriptor, PEOE_PC-_#, increased (Figure 7).
The lowest AIC (Table 5) suggested that the multiregression equation for fish toxicity
Downloaded by [FU Berlin] at 01:11 07 May 2015

(model F-II) and the simple linear regression equation for Daphnia toxicity (model D-I)
were the best models for the CH-group dataset. The other statistical data, namely, the
highest r2adj , the lowest SE, and the highest q2LOO among the three models (F-I, II, III or D-
I, II, III), also supported the robustness and goodness-of-fit of F-II and D-I. Under the
linear regression framework, the correlation between toxicity and PEOE_PC-_# was better
than that between toxicity and log POW. The plot of the correlation between measured and
calculated toxicities for F-III showed that the calculated fish toxicities of three acrylic acids
(whose measured values were less than 0.5 unit of log (1/LC50)) were overestimated, and
the correlation plot for D-III showed that the Daphnia toxicity was independent of the log
POW descriptor (Figure 8). In addition, Equations (1) and (2) had a smaller number of
reference data than the equations for models F-II (or F-I) and D-I, which had larger r2 and
q2LOO values (Table 5) than the values for Equations (1) and (2). The combination of
Acrylates and Acids acrylic in the KATE2011 classification with the PEOE descriptor
enhanced the correlation between measured and calculated values, the robustness of the
QSAR models, and the number of reference data.
The fish toxicity of 20 calculated by model F-I (the PEOE descriptor only) was the
most underestimated value for the CH-group compounds: the measured log(1/LC50) was
1.80, whereas the calculated value was 0.96 (Table 2 and Appendix 5 (available via the
Supplementary Content tab on the article’s online page at http://dx.doi.org/10.1080/
1062936X.2012.719542), respectively). The three carbonyl groups in the ring structure of
20 enhanced the magnitude of the negative charges of 20, and the electrostatic effect of
these substructures led to the underestimate of the fish toxicity. In contrast, the Daphnia
toxicity of 20 was well calculated by the model: the measured log(1/EC50) was 0.69, and the
calculated value was 0.71 (Table 2 and Appendix 5, respectively). When 20 was treated as
an outlier in the fish toxicity data, the goodness-of-fit and robustness of the simple linear
regression equation improved slightly (see Appendix 6, available via the Supplementary
Content tab on the article’s online page at http://dx.doi.org/10.1080/1062936X.2012.
719542); r2adj increased from 0.67 to 0.76, and RMSE [SE] decreased from 0.44 [0.47] to
0.39 [0.42]. However, even when 20 was omitted, the AIC for the simple linear regression
analysis involving the PEOE descriptor (22.5) was still higher than the AIC for the
multiregression analysis (24.9). When 11, which is neither an acrylate nor an acrylic acid,
was treated as an outlier, the goodness-of-fit and robustness of the simple linear regression
 SAR and QSAR in Environmental Research 743

Table 5. QSAR equations for correlation between fish toxicity, log(1/LC50 [mmol L1]), or Daphnia
toxicity, log(1/EC50 [mmol L1]), and PEOE_PC-_# or log POW for CH-group compounds (11–25).

RMSE RMSE_LOO
Model AIC r2 (r2adj ) [SE] q2LOO [SELOO]

Fish toxicity: n ¼ 15
F-I log(1/LC50) ¼ 3.03(PEOE_PC-_#) 20.7 0.70 (0.67) 0.44 [0.47] 0.59 0.51 [0.55]
þ 4.45
F-II log(1/LC50) ¼ 2.45(PEOE_PC-_#) 25.2 0.80 (0.77) 0.35 [0.39] 0.70 0.44 [0.49]
þ 0.25(log POW) þ 3.47
F-III log(1/LC50) ¼ 0.45(log POW) þ 0.77 11.4 0.44 (0.39) 0.60 [0.64] 0.30 0.68 [0.73]

Daphnia toxicity: n ¼ 10
D-I log(1/EC50) ¼ 2.09(PEOE_PC-_#) 24.4 0.82 (0.79) 0.24 [0.27] 0.73 0.30 [0.33]
þ 3.12
D-II log(1/EC50) ¼ 2.07(PEOE_PC-_#) 22.4 0.82 (0.76) 0.24 [0.29] 0.73 0.30 [0.36]
Downloaded by [FU Berlin] at 01:11 07 May 2015

þ 0.02(log POW) þ 3.08

D-III log(1/EC50) ¼ 0.16(log POW) þ 0.82 8.14 0.07 (0.05) 0.54 [0.61] 0.18 0.71 [0.79]

Notes: PEOE_PC-_# denotes PEOE_PC- divided by the number of ,-unsaturated carbonyl

oxygen atoms in the compound. PEOE_PC- ¼ the sum of the PEOE negative atomic charges. n, r2
(r2adj ), RMSE [SE], q2LOO , RMSE_LOO [SELOO], and Outlier denote number of chemicals, squared
correlation coefficient (with adjustment for the degrees of freedom), root mean square error
[standard error], leave-one-out version of the squared correlation coefficient, leave-one-out version of
RMSE [SE], and outliers for the regression equations, respectively. Underlining indicates data for
models with the lowest AIC for the fish or Daphnia toxicity.

analysis decreased, but the multiregression analysis gave the lowest AIC (see Appendix 6).
Therefore, neither 11 nor 20 was a factor in explaining why the multiregression analysis
was better than the simple linear regression analysis for predicting the fish toxicity of
acrylates and acrylic acids. Treating compounds 11 and 20 as outliers in the QSAR models
for the CH-group compounds was unnecessary from a statistical point of view.
In summary, PEOE_PC-_# was a practical descriptor for expressing both fish and
Daphnia toxicities of acrylic acids and compounds with acrylate-like substructures, that is,
,-unsaturated esters bearing a hydrogen atom on the  carbon.

3.3 Applicability of the constructed QSAR models to the CH3-group compounds

We next confirmed the applicability of the QSAR models described in section 3.2; that is,
we determined whether the models based on PEOE_PC-_# for the CH-group compounds
were practical for prediction of the toxicities of other ,-unsaturated esters.
We evaluated the values calculated by models F-I, F-II, and D-I for the CH3-group
compounds (Figure 9). These three models were simple linear regressions involving the
PEOE descriptor or had the lowest AIC values among the fish LC50 and Daphnia EC50
models in Table 5. The fish and Daphnia toxicities for the CH3-group compounds
calculated by using the models in Table 5 are listed in Appendix 5.
Two of the CH3-group compounds, 28 and 35, are not methacrylates; therefore, the
calculated values for these compounds tended to overestimate both the fish and the
Daphnia toxicities by nearly 1.0 unit. Although the Daphnia toxicity of 35 seemed to be
 744 A. Furuhama et al.
Downloaded by [FU Berlin] at 01:11 07 May 2015

Figure 7. Correlation between fish (a) and Daphnia (b) toxicities and PEOE_PC-_# for the CH-
group compounds.

calculated well, the measured log(1/EC50) of 35 should be lower than the value
(EC50 4 10 mg L1) in the plot shown in Figure 9 for model D-1.
The other 13 CH3-group compounds are methacrylates, which fall into the Michael
addition domain, and have substructures similar to those of the CH-group compounds,
except that the former have a methyl group attached to the  carbon. The calculated
toxicities of 26 (both fish and Daphnia), 34 (fish), 37 (Daphnia), and 38 (both fish and
Daphnia) were overestimated by 1.0 unit, and the toxicities of the other compounds were
within 1.0 unit of the measured values (Figure 9). For 26, 34, and 38, toxicities calculated
from models F-I and F-II differed by only 0.21 units or less. Interestingly, both the fish and
the Daphnia toxicities of 26 and 38 were overestimated by 1.0 unit when the PEOE
descriptor models were used, because the slope of the plot of PEOE_PC-_# for models F-I
and D-I was positive, but the PEOE_PC-_# values of 26 and 38 were not negatively shifted
enough to bring the calculated values downs as low as the measured values. These results
suggest that PEOE_PC-_# is unsuitable for methacrylates with 2-(diethylamino)ethyl and
2-hydroxyethyl groups. First, the PEOE_PC-_# of 2(diethylamino)ethyl methacrylate 26
(1.06) was slightly smaller than the values of 2 -(dimethylamino)ethyl methacrylate 32
(1.03) and 2 -(dimethylamino)ethyl acrylate 17 (1.00). Both 32 and 17 showed much
higher measured toxicities than 26. Second, the PEOE_PC-_# of 2-hydroxyethyl
 SAR and QSAR in Environmental Research 745
Downloaded by [FU Berlin] at 01:11 07 May 2015

Figure 8. Correlation between measured and calculated fish (upper) and Daphnia (lower) toxicities
for the CH-group compounds. The labels and reference data in the correlations correspond to the
QSAR models in Table 5.

methacrylate 38 (1.07) was only slightly different from that of 2-hydroxyethyl acrylate 23
(1.06), which had a measured toxicity much higher than that of 38. Moreover, methyl
methacrylate 34 tended to have relatively smaller PEOE_PC-_# value than methyl acrylate
24 (0.67 for 24 and 0.77 for 34), and the Daphnia toxicity of 34 may be overestimated
by models based on PEOE_PC-_# values for compounds with acrylate-like substructures.
We concluded that the QSAR models were not adequate to extend the applicability
domain to the CH3-group compounds.

4. Conclusions
To predict acute fish and Daphnia toxicities, we used Gasteiger’s PEOE and log POW to
construct equations including a 2D descriptor. We first determined how the electrostatic
effects of the Mulliken atomic charge on the carbonyl oxygen affected fish toxicity
predictions, and then we applied the descriptors based on the PEOE partial charges to
,-unsaturated ketones and aldehydes (A-group compounds) and to acrylic acids and
compounds with acrylate-like substructures (CH-group compounds). The QSAR models
for the fish and Daphnia toxicities of the CH-group compounds constructed with the
PEOE descriptor permitted us to include electrostatic effects effectively. Then we
evaluated the applicability of the constructed QSAR models (F-I, F-II and D-I in Table 5)
 746 A. Furuhama et al.
Downloaded by [FU Berlin] at 01:11 07 May 2015

Figure 9. Applicability of QSAR models F-I, F-II, and D-I in Table 5, as indicated by
the correlation between the measured and calculated toxicities for the CH3-group compounds.
The labels correspond to the models in Table 5. The measured fish toxicities of 24 and 28 and
the measured Daphnia toxicity of 35 were taken from, respectively, the lowest LC50 (100 mg L1) and
EC50 (10 mg L1) values (Table 3) and are indicated by crosses.

to methacrylates and compounds with similar substructures (CH3-group compounds). The

fish and Daphnia toxicities of some of the CH3-group compounds were underestimated by
the QSAR models.
We concluded that the QSAR models based on the PEOE descriptor, sometimes
combined with a hydrophobicity descriptor, were useful for predicting fish LC50 and
Daphnia EC50 values for ,-unsaturated esters with an  hydrogen, as well as for acrylates
and acrylic acids. Because the PEOE descriptor can be used to estimate the electrostatic
effects based on atom types and bond orders (i.e. 2D structures), QSAR models based on
this descriptor can be expected to be applicable for the screening of chemicals.

Supplementary materials
The structures of the CH- and CH3-group compounds, statistical information, and
toxicities of target compounds calculated by models F-I, II, and III and D-I, II, and III are
 SAR and QSAR in Environmental Research 747

available as appendixes, available via the Supplementary Content tab on the article’s
online page at http://dx.doi.org/10.1080/1062936X.2012.719542.

References

[1] OECD, OECD Environment Monographs No. 58, Report of the OECD workshop on quantitative
structure activity relationships (QSARs) in aquatic effects assessment, Paris, 1992. Available at
http://www.oecd.org/dataoecd/19/27/35220377.pdf
[2] T.I. Netzeva, M. Pavan, and A.P. Worth, Review of (quantitative) structure-activity relationships
for acute aquatic toxicity, QSAR Comb. Sci. 27 (2008), pp. 77–90.
[3] A. Worth, T.I. Netzeva, and G. Patlewicz, 10. Predicting toxicological and ecotoxicological
endpoints, in Risk assessment of chemicals: an introduction, F.W. Van Leeuwen and
T.G. Vermeire, eds., Springer-Verlag, The Netherlands, Dordrecht, 2007, pp. 576–578.
[4] OECD, OECD Environment health and safety publications series on testing and assessment No.
58, Report on the regulatory uses and applications in OECD member countries of (quantitative)
structure-activity relationship [(Q)SAR] models in the assessment of new and existing chemicals,
Downloaded by [FU Berlin] at 01:11 07 May 2015

Paris, 2006. Available at http://www.oecd.org/dataoecd/55/22/38131728.pdf

[5] OECD, OECD Environment health and safety publications series on testing and assessment No.
69, Guidance document on the validation of (quantitative) structure-activity relationships
[(Q)SAR] models, Paris, 2007. Available at http://www.oecd.org/dataoecd/55/35/38130292.pdf
[6] OECD, OECD Quantitative structure-activity relationships project [(Q)SARs], Paris. Available
at http://www.oecd.org/env/hazard/qsar
[7] ECOSAR. Available at http://www.epa.gov/oppt/newchems/tools/21ecosar.htm
[8] H.J.M. Verhaar, C.J. Vanleeuwen, and J.L.M. Hermens, Classifying environmental pollutants. 1:
Structure-activity relationships for prediction of aquatic toxicity, Chemosphere 25 (1992),
pp. 471–491.
[9] KATE, KAshinou Tool for Ecotoxicity (KATE) is an ecotoxicity QSAR modeling system
researched and developed by the Research Center for Environmental Risk at the National
Institute for Environmental Studies under contract from the Ministry of the Environment from
FY 2004 to FY2010. Available at http://kate.nies.go.jp
[10] A. Furuhama, T. Toida, N. Nishikawa, Y. Aoki, Y. Yoshioka, and H. Shiraishi, Development of
an ecotoxicity QSAR model for the KAshinhou Tool for Ecotoxicity (KATE) system, March 2009
version, SAR QSAR Environ. Res. 21 (2010), pp. 403–413.
[11] A. Furuhama, K. Hasunuma, Y. Aoki, Y. Yoshioka, and H. Shiraishi, Application of chemical
reaction mechanistic domains to an ecotoxicity QSAR model, KAshinhou Tool for Ecotoxicity
(KATE), SAR QSAR Environ. Res. 22 (2011), pp. 505–523.
[12] Ministry of the Environment in Japan, Results of eco-toxicity tests data conducted by Ministry of
the Environment in Japan. Available at http://www.env.go.jp/chemi/sesaku/02e.pdf
[13] US EPA, US Environmental Protection Agency’s fathead minnow database. Available at http://
www.epa.gov/med/Prods_Pubs/fathead_minnow.htm
[14] S. Dimitrov, T. Pavlov, G. Veith, and O. Mekenyan, Simulation of chemical metabolism for fate
and hazard assessment. I. Approach for simulating metabolism, SAR QSAR Environ. Res. 22
(2011), pp. 699–718.
[15] S. Dimitrov, T. Pavlov, N. Dimitrova, D. Georgieva, D. Nedelcheva, A. Kesova, R. Vasilev,
and O. Mekenyan, Simulation of chemical metabolism for fate and hazard assessment. II
CATALOGIC simulation of abiotic and microbial degradation, SAR QSAR Environ. Res. 22
(2011), pp. 719–755.
[16] S. Dimitrov, N. Dimitrova, D. Georgieva, K. Vasilev, T. Hatfield, J. Straka, and O. Mekenyan,
Simulation of chemical metabolism for fate and hazard assessment. III. New developments of the
bioconcentration factor base-line model, SAR QSAR Environ. Res. 23 (2012), pp. 17–36.
 748 A. Furuhama et al.

[17] S. Dimitrov, N. Dimitrova, T. Parkerton, M. Comber, M. Bonnell, and O. Mekenyan, Base-line

model for identifying the bioaccumulation potential of chemicals, SAR QSAR Environ. Res. 16
(2005), pp. 531–554.
[18] S.D. Dimitrov, N.C. Dimitrova, J.D. Walker, G.D. Veith, and O.G. Mekenyan,
Bioconcentration potential predictions based on molecular attributes – an early warning
approach for chemicals found in humans, birds, fish and wildlife, QSAR Comb. Sci. 22 (2003),
pp. 58–68.
[19] M. Moosus and U. Maran, Quantitative structure–activity relationship analysis of acute toxicity
of diverse chemicals to Daphnia magna with whole molecule descriptors, SAR QSAR Environ.
Res. 22 (2011), pp. 757–774.
[20] A. Furuhama, Y. Aoki, and H. Shiraishi, Consideration of reactivity to acute fish toxicity of ,-
unsaturated carbonyl ketones and aldehydes, SAR QSAR Environ. Res. 23 (2012), pp. 169–184.
[21] J.A.H. Schwöbel, Y.K. Koleva, S.J. Enoch, F. Bajot, M. Hewitt, J.C. Madden, D.W. Roberts,
T.W. Schutz, and M.T.D. Cronin, Measurement and estimation of electrophilic reactivity for
predictive toxicology, Chem. Rev. 111 (2011), pp. 2562–2596.
[22] T.W. Schultz, T.I. Netzeva, D.W. Roberts, and M.T.D. Cronin, Structure–toxicity relationships
for the effects to Tetrahymena pyriformis of aliphatic, carbonyl-containing, alpha, beta-
unsaturated chemicals, Chem. Res. Toxicol. 18 (2005), pp. 330–341.
Downloaded by [FU Berlin] at 01:11 07 May 2015

[23] J.W. Yarbrough and T.W. Schultz, Abiotic sulfhydryl reactivity: A predictor of aquatic toxicity
for carbonyl-containing alpha, beta-unsaturated compounds, Chem. Res. Toxicol. 20 (2007),
pp. 558–562.
[24] D.W. Roberts and A. Natsch, High throughput kinetic profiling approach for covalent binding to
peptides: Application to skin sensitization potency of Michael acceptor electrophiles, Chem. Res.
Toxicol. 22 (2009), pp. 592–603.
[25] J.A.H. Schwöbel, D. Wondrousch, Y.K. Koleva, J.C. Madden, M.T.D. Cronin, and
G. Schüürmann, Prediction of Michael-type acceptor reactivity toward glutathione, Chem. Res.
Toxicol. 23 (2010), pp. 1576–1585.
[26] J. Gasteiger and M. Marsili, Iterative partial equalization of orbital electronegativity – a rapid
access to atomic charges, Tetrahedron 36 (1980), pp. 3219–3228.
[27] C.L. Russom, S.P. Bradbury, S.J. Broderius, D.E. Hammermeister, and R.A. Drummond,
Predicting modes of toxic action from chemical structure: Acute toxicity in the fathead minnow
(Pimephales promelas), Environ. Toxicol. Chem. 16 (1997), pp. 948–967.
[28] Molecular Operating Environment (MOE), 2008.1002, Chemical Computing Group Inc., 1010
Sherbooke St. West, Suite #910, Montreal, QC, Canada H3A 2R7, 2008.
[29] M.J. Frisch, G.W. Trucks, H.B. Schlegel, G.E. Scuseria, M.A. Robb, J.R. Cheeseman,
G. Scalmani, V. Barone, B. Mennucci, G.A. Petersson, H. Nakatsuji, M. Caricato, X. Li,
H.P. Hratchian, A.F. Izmaylov, J. Bloino, G. Zheng, J.L. Sonnenberg, M. Hada, M. Ehara,
K. Toyota, R. Fukuda, J. Hasegawa, M. Ishida, T. Nakajima, Y. Honda, O. Kitao, H. Nakai,
T. Vreven, J.A. Montgomery Jr, J.E. Peralta, F. Ogliaro, M. Bearpark, J.J. Heyd, E. Brothers,
K.N. Kudin, V.N. Staroverov, R. Kobayashi, J. Normand, K. Raghavachari, A. Rendell,
J.C. Burant, S.S. Iyengar, J. Tomasi, M. Cossi, N. Rega, J.M. Millam, M. Klene, J.E. Knox,
J.B. Cross, V. Bakken, C. Adamo, J. Jaramillo, R. Gomperts, R.E. Stratmann, O. Yazyev,
A.J. Austin, R. Cammi, C. Pomelli, J.W. Ochterski, R.L. Martin, K. Morokuma,
V.G. Zakrzewski, G.A. Voth, P. Salvador, J.J. Dannenberg, S. Dapprich, A.D. Daniels,
O. Farkas, J.B. Foresman, J.V. Ortiz, J. Cioslowski, and D.J. Fox, Gaussian 09, Revision A.02,
Gaussian, Inc., Wallingford CT, 2009.
[30] J. Sadowski, J. Gasteiger, and G. Klebe, Comparison of automatic 3-dimensional model builders
using 639 X-Ray structures, J. Chem. Inf. Comput. Sci. 34 (1994), pp. 1000–1008.
[31] CORINA, The 3D structure generator CORINA. Information is available at http://
www.molecular-networks.com, Molecular Networks GmbH, Erlangen, Germany.
[32] D. Weininger, SMILES, A chemical language and information-system. 1. Introduction to
methodology and encoding rules, J. Chem. Inf. Comput. Sci. 28 (1988), pp. 31–36.
 SAR and QSAR in Environmental Research 749

[33] PEOE_PC- in MOE. Available at http://www.chemcomp.com/journal/descr.htm#Charge

[34] F. Jensen, Wave function analysis, in Introduction to computational chemistry, John Wiley & Sons
Ltd, Chichester, 2007, pp. 293–314.
[35] J. Gasteiger, Empirical approaches to the calculation of properties, in Chemoinformatics: A
Textbook, J. Gasteiger and T. Engel, eds., WILEY-VHC GmbH & Co, KGaA, Weinheim, 2003,
pp. 320–337.
[36] R Development Core Team, R: A Language and Environment for Statistical Computing, The
R foundation for statistical Computing, Vienna, Austria, 2008.
[37] M.V. Putz, On absolute aromaticity within electronegativity and chemical hardness reactivity
pictures, MARCH 64 (2010), pp. 391–418.
[38] M.V. Putz, Systematic formulations for electronegativity and hardness and their atomic scales
within density functional softness theory, Int. J. Quantum. Chem. 106 (2006), pp. 361–389.
[39] S.A. Chicu and M.V. Putz, Koln-Timisoara molecular activity combined models toward
interspecies toxicity assessment, Int. J. Mol. Sci. 10 (2009), pp. 4474–4497.
[40] ClogP, See the CLOGP reference manual at http://www.daylight.com/dayhtml/doc/clogp/
index.html. The underlying program of CLOGP is copyrighted by Pomona College and BioByte,
Inc. (Claremont, CA).
Downloaded by [FU Berlin] at 01:11 07 May 2015