1602
Chem. Res. Toxicol. 2002, 15, 1602-1609
Structure-Toxicity Relationships for Aliphatic Chemicals Evaluated with Tetrahymena pyriformis T. Wayne Schultz,*,† Mark T. D. Cronin,‡ Tatiana I. Netzeva,‡ and Aynur O. Aptula§ The University of Tennessee, College of Veterinary Medicine, 2407 River Drive, Knoxville, Tennessee 37996-4543, School of Pharmacy and Chemistry, Liverpool John Moores University, Byrom Street, Liverpool, L3 3AF, England, and Department of Chemical Ecotoxicology, UFZ Centre for Environmental Research, Leipzig-Halle, Permoserstrasse 15, D-04318, Germany Received August 1, 2002
Quantitative structure-activity relationships were developed for the toxicity data of 500 aliphatic chemicals tested in the two-day Tetrahymena pyriformis population growth impairment assay. These chemicals represented a number of structural classes spanning a variety of mechanisms of toxic action including narcoses and electrophilic mechanisms. A series of quantitative structure-toxicity models correlating toxic potency [log(IGC50-1)] with a limited number of mechanistically interpretable descriptors were developed for toxicological domains within the data set. The descriptors included the 1-octanol/water partition coefficient (log Kow) (for hydrophobicity) and the energy of the lowest unoccupied molecular orbital (Elumo) to quantify electrophilic reactivity. Neutral (nonpolar) narcosis was well modeled by the equation [log(IGC50-1) ) 0.723(0.140) (log Kow) - 1.79(0.031); n ) 215, r2 (adj.) ) 0.926, s ) 0.274, r2 (pred.) ) 0.925]. Chemical classes fitting this domain included saturated alcohols, ketones, nitriles, esters, and sulfur-containing compounds. When the neutral narcotic chemicals were combined with diester narcotics, carboxylic sodium salts, Schiff-based forming aldehydes, electrophilic compounds capable of acting by a SN2 mechanism, and proelectrophiles, the model [log(IGC50-1) ) 0.45(0.014) (log Kow) - 0.342 (0.035) (Elumo) - 1.11(0.05); n ) 353, r2 (adj.) ) 0.859, s ) 0.353, r2 (pred.) ) 0.857] provided a good fit to the data. The model [log(IGC50-1) ) 0.273(0.018) (log Kow) - 0.116(0.056) (Elumo) - 0.558(0.054); n ) 35, r2 (adj.) ) 0.873, s ) 0.141, r2 (pred.) ) 0.838] provided an excellent fit of the data for compounds containing a carboxyl [RC(dO)O] group. The toxicity of aliphatic amines [RCN] was modeled by the equation [log(IGC50-1) ) 0.676(0.048) (log Kow) - 1.23(0.08) n ) 30, r2 (adj.) ) 0.873, s ) 0. 336, r2 (pred.) ) 0.848]. The potency of saturated aliphatic isothiocyanates was a constant (0.0202 mM). Aliphatic chemicals that did not model well by equations involving log Kow and Elumo included amino alcohols and R-haloactivated compounds.
Introduction Among the most prevalent industrial organic chemicals in the world (as defined by the high production volume chemical list) are a variety of aliphatic acids, alcohols, esters, saturated and unsaturated alkanes, and halogenated alkanes (1). There is an increased emphasis in predicting the hazardous effects of these chemicals from molecular structure (2). The ability to use a quantitative structure-activity relationship (QSAR)1 to estimate accurately the relative ecotoxicity of such a diverse group of chemicals would be of value to industry and regulatory bodies alike. To develop such a toxicity-based QSAR requires three componentssan effects data set, which * To whom correspondence should be addressed. E-mail: tschultz@ utk.edu. Phone: (865) 974-5826. Fax: (865) 974-5640. † The University of Tennessee. ‡ Liverpool John Moores University. § UFZ Centre for Environmental Research. 1 Abbrevations: QSARs, quantitative structure-activity relationships; IGC50, 50% growth inhibitory concentration; log Kow, logarithm of 1-octanol/water partition coefficient; Elumo, energy of the lowest unoccupied molecular orbital; Ehomo, energy of the highest occupied molecular orbital; Smax, maximum superdelocalizability; ElipVol, ellipsoidal volume; 3χC, connectivity cluster index of third order; 3χvp, connectivity valence path index of third order; log H, logarithm of Henry’s law constant.
provides a uniform measure of the toxicity for a group of chemicals, molecular structure/property data (i.e., descriptors) for each chemical within the group, and a statistical method with which to form a relationship between toxicity and structure. A variety of ecotoxicity data sets have been compiled for QSAR analysis (3). Of these the population growth inhibition of the freshwater ciliate Tetrahymena pyriformis, which has been derived for the express purposes of QSAR development and validation, is among the largest. This data set includes toxicity data for a large number of aliphatic organic substances. These data originated from a single research facility using a standard test protocol and include chemicals that vary widely in molecular structures and potential mechanisms of toxic action (see refs 3 and 4). Recent discussions have focused on a number of essential or desirable characteristics for ecotoxicological QSARs (5). Included among these characteristics was the need for easily interpretable (i.e., transparent) QSARs based on a limited number of mechanistically interpretable molecular descriptors with defined chemical domains (5). As such, the appropriate use of molecular descriptors is required for successful QSAR development and ap-
10.1021/tx025589p CCC: $22.00 © 2002 American Chemical Society Published on Web 11/16/2002
QSARs for Aliphatic Chemicals
plication. This includes not only the use of an appropriate number of descriptors for the size and complexity of the data set, but also the use of descriptors based on fundamental molecular properties. The properties of compounds are related to their molecular structure. Whereas chemicals are normally thought of in a twodimensional (D) structure, toxicity is a manifestation of the 3-D structure of a molecule. Properties are also typically manifestations of 3-D structure. A variety of properties have been used in structure-toxicity modeling including physicochemical and quantum chemical properties. Physicochemical properties include descriptors for the hydrophobic, electronic, and steric properties of a molecule (6). Quantum chemical properties include charge and energy values. Certain toxicological effects may be easily related to physicochemical properties. For instance, the uptake of most industrial organic chemicals to the site of action is by passive diffusion and is best modeled by hydrophobicity, most often quantified by the 1-octanol/water partition coefficient (log Kow) (7). Interaction of the chemical with the site of action is more complicated and is quantified by a number of molecular parameters that typically describe electronic and/or steric properties (8). The most descriptive interactive parameters appear to be quantum chemical parameters, especially ones based on frontier orbitals (9). The studies of Karabunarliev et al. (9) also suggest that for modeling purposes segregation of compounds by specific electrophilic mechanism or chemical domain is required to provide high quality predictions. This is in large part because different descriptors are required for different QSARs and the values of some descriptors, especially the quantum chemical ones, are impacted by whether the chemical is aliphatic or aromatic. However, selection of the correct domain for toxicity prediction may not be an easy task (10). This is demonstrated by the problems of modeling the toxicity to fish of R,β-unsaturated compounds (9). The objective of this investigation was to develop a series of chemical class-based QSARs describing the toxicity to T. pyriformis that are significant, both in terms of numbers and chemical heterogeneity for a data set of aliphatic compounds. QSARs were developed using regression analysis and a limited number of mechanistically interpretable molecular descriptors. The chemical classes represented in the analyses included saturated and unsaturated alcohols (including R-unsaturated compounds), diols, saturated and unsaturated esters, diesters, ketones, diketones, carboxylic acids and dicarboxylic acids, amines, aminolalcohols, cyanides, dicyanides, strained-ring lactones, a variety of halogenated compounds including hydrocarbons, alcohols, esters, cyanides and esters, and small groups of the sodium salts of carboxylic acids, oximes, carbamates, hydrazides, sulfurcontaining compounds, and isothiocyanates.
Materials and Methods Chemicals. The toxicity of 500 aliphatic molecules representing several chemical classes but excluding R,β-unsaturated compounds was evaluated. Caution: The following chemicals are hazardous and should be handled carefully. As reported here several of these chemicals have significant acute toxicity. In addition, a number of compounds are potential mutagens and skin sensitizers (11). The molecules were obtained commercially (Aldrich Chemical Co., Milwaukee, WI; MTM Research Chemicals or Lancaster Synthesis Inc., Windham, NH). In the vast majority of cases purity was >95%. In all cases no further
Chem. Res. Toxicol., Vol. 15, No. 12, 2002 1603 purification was undertaken. Biological Data. Population growth impairment testing with the common ciliate T. pyriformis (strain GL-C) was conducted following the protocol described by (4). This 40-h assay is static in design and uses population density quantified spectrophotometrically at 540 nm as its endpoint. The test protocol allows for eight to nine cell cycles in controls. Following range finding, each chemical was tested in three replicate tests (or assays). Two controls were used to provide a measure of the acceptability of the test by indicating the suitability of the medium and test conditions as well as a basis for interpreting data from other treatments. The first control had no test material and was inoculated with T. pyriformis. The other, a blank, had neither test material nor ciliates. Each test replicate consisted of six to ten different concentrations of each test material with duplicate flasks of each concentration. Only replicates with controlabsorbency values of >0.60 but of 5) have insufficient water solubility in the test system to bring about 50% lethality. These limits in log Kow define the domain of applicability in terms of hydrophobicity, not only for the test system, but also for the QSARs developed. Further, the upper limits of log Kow are consistent with the suggested practical capabilities of its measurement, e.g., for the shake flask (30). In addition, there is evidence that for very hydrophobic chemicals that there may be a change in mechanism of action, especially if for specific toxicants that may act in the aqueous compartments of the cell (31). Recent efforts (27-29) have modeled the toxicity of aliphatic compounds across chemical classes and molecular mechanisms. However, those previous studies were based on much more structurally limited groups of chemicals than considered in this study. Equation 14 presented in this study suggests strongly that, at least for selected chemical domains, a the two-parameter model provides a means of predicting toxicity spanning mechanisms that include neutral narcosis, nonspecific soft electrophilicity, and the specific electrophilic mechanisms of Schiff-base formation and bimolecular nucleophilic substitution. A number of chemical classes were not modeled well by eq 14. Such classes should be considered to reside outside the domain of the model. One of the most important classes considered to reside outside the domain of eq 14 was the carboxylic acids (including R,β-unsaturated acids). In addition to the coefficients for both hydrophobicity and electrophilicity being different in eqs 14 and 15, it was considered that the rationale for excluding carboxylic acids from the general model (i.e., eq 14) was due to the contribution of ionized species to toxicity (32). Moreover, since pH effects ciliate generation time (33) and carboxylic acids affect the buffering capacity of the growth medium, the capacity of the system to withstand increased acidity without significantly altering generation time is concentration dependent. The primary amines were also poorly modeled by both eqs 1 and 14. The toxicity of aliphatic amines has been examined in a number of aquatic assays (see ref 34). These structure-toxicity studies, which include data for fish, microcrustaceans, protozoa, and algae, show amine narcotic toxicity to be in excess of baseline narcosis (e.g., eq 1). However, no fish acute toxic response syndrome similar to that for neutral narcosis (35) has been reported for amines. A comparison of eq 1 with eq 17 shows a similar slope for the Kow dependence but a significantly
Chem. Res. Toxicol., Vol. 15, No. 12, 2002 1607
greater intercept for amines. At the pH of the test system, amines with an ionization constant (pKa) greater than 9.0 are protonated. It is known that protonated amines have a higher affinity for membranes than nonionized compounds (36); therefore, the toxicity of such compounds would be increased over that expected from nonpolar narcosis. As such, it is not yet possible to determine if a separate mechanism of action exists for the amines or whether their requirements for separate modeling is merely an artifact of poorly calculated log Kow values. The isothiocyanates were a further group of chemicals that were not modeled well by eq 14. Schultz and Comeaux (19) noted their toxicity is independent of log Kow. While the log(IGC50-1) of saturated aliphatic isothiocyanates was a constant (i.e., 0.0202 ( 0.0023 mM), a comparison of potency of 1,3-propylene diisothiocyanate with n-propyl isothiocyanate reveals that the introduction of a second isothiocyanate group sharply increased toxicity (19). Moreover, a comparison of allyl isothiocyanate and n-propyl isothiocyanate revealed a decrease in hydrogen saturation also results in an increase in toxicity (19). An additional descriptor to log Kow and Elumo was required to model of the toxicity of the pro-electrophiles. It was found that the third order cluster molecular connectivity index 3χC, which clearly accounts for the branching in the molecules (37), improves the initial twodescriptor model. For this particular subset, the positive coefficient of 3χC indicates that toxicity is greater when the hydroxyl group in R-position to the triple bond is associated to a secondary carbon atom and lower when it is attached to a primary one. Interestingly, the usage of molecular cluster connectivity together with Kow and Elumo was shown also to be beneficial in the modeling of hepatotoxic effects of various aliphatic alcohols in perfused rat liver model (38). The modeling of the toxicity to Tetrahymena of so-called proelectrophiles has been a problem (3), especially as related to the explanation provide for the toxicity of these substances to fish (17, 23). The R-haloactivated compounds were also not well modeled by eq 14. Their toxicity appears to be independent of hydrophobicity (see eq 18). Interestingly, three descriptorssElumo, the maximum superdelocalizability (Amax), and the ellipsoidal volume (ElipVol)swere found to be important to model toxic potency of the compounds undergoing R-haloactivation. The chemical structure of these compounds, i.e., the presence of polarized double or triple bond next to the halogenated carbon atom, determines their high reactivity. The high reactivity is indicated by a high value for electrophilicity. The reactivity in Phase II enzymes is mediated by a SN2-type transition state with the partially negative charged sulfur atom from the thiol groups of glutathione S-transferases (39). Consequently, it is not surprising that their toxicity is correlated well with electrophilicity descriptors Elumo and Amax. The former, Elumo, reflects the reactivity of the whole molecule and depends mainly on the number and type of the halogen atoms, as well as on the type of the polarized group, while the latter (Amax) depends more on the arrangement of the heteroatoms. The third descriptor, ElipVol, appears out of place in this equation, as volume reflects the molecular bulk. However, when the molecule is nonspherical in shape, there is less similarity between ElipVol and molecular volume. Thus, for the aliphatic compounds considered, ElipVol accounts more
1608
Chem. Res. Toxicol., Vol. 15, No. 12, 2002
for shape, rather than the bulk, of the molecules. The positive coefficient of the descriptor suggests higher toxicity for the longer and slimmer (more linear), and lower toxicity for the shorter and more bulky (less linear), molecules. This finding could be pertinent to the transport of the molecules through the cell membrane via passive diffusion, which is generally easier for the more linear molecules. The three-parameter model eq 19, obtained for the amino alcohols, reveals the totally different nature of the structure-toxicity relationship of these compounds compared to other chemicals considered in this study. Amino alcohols, especially β-hydroxylamines, are corrosive. Their toxicity was found to be independent of hydrophobicity. To describe reactivity, the energy of the highest occupied molecular orbital was found to be significant, thus suggesting nucleophilic rather than electrophilic type of reactivity. The connectivity index 3χvp is sensitive both to the size and branching of the molecules. It is higher for molecules which have substituents in an R,β-position to each other, such as 2-amino-3-methyl-1-pentanol and 2-amino-3,3-dimethyl-butanol. The third descriptor in eq 19, log H, accounts for volatility, predicting higher toxicity for less volatile chemicals. It should be noted that the presence of at least two polar groups in the amino alcohols resulted in very hydrophilic compounds. A further nine compounds were not included in the QSAR analysis as they fell outside the general cutoff for hydrophilicity (i.e., they had log Kow values less than -1) and hence only low confidence could be assigned to their toxicity value. As the use of structure-toxicological relationships proceeds into the realm of risk assessment, the need for a more thorough understanding of the practicalities of a QSAR is increasing. The specific issues of quality, transparency, and more specific domain identification have been recognized (5). However, determining the quality of a QSAR is often a difficult task that must be extended beyond a high coefficient of determination or the use of additional descriptors to increase statistical fit. For instance, a high quality QSAR can only be constructed and validated with high quality data. Transparency is also a critical issue for the acceptance and wider use of QSARs (40). There are a number of elements to assessing the transparency of a QSAR. Initially, for a QSAR to be termed transparent the data set on which it based should be openly available for review. A transparent QSAR should also be developed with descriptors that quantify the critical aspects of toxicity and thus are mechanistic-based and easily interpreted. It should be noted that the use of mechanism-based descriptors differs from the QSAR being based on a mechanism of toxic action. Transparency also relates to the information obtainable from the statistics, which ranges from difficult to interpret neural networks algorithms to easily interpretable multiple linear regression. The chemical domain of a QSAR is typically established by outliers. As noted by Egan and Morgan (41), outliers from a QSAR are molecules that do not fit the model or are poorly predicted by it. While there may be several potential reasons for a chemical being an outlier, most often outliers are explained as acting by a different mechanism of action from the chemicals, which are well modeled by the QSAR. In conclusion, the largest consistent and reliable available database for the prediction of the toxicity of aliphatic compounds has been presented. Developing QSARs on a
Schultz et al.
mechanism of action approach results in a high quality, hydrophobicity-dependent model for neutral narcosis. The aliphatic amines are not well modeled by the general narcosis model, the reason for which was not clear, but may be due to erroneous calculations of log Kow. A more global two-parameter model for the prediction of aliphatic compounds, which encompasses a greater number of mechanisms of action, is presented. The toxicity of a number of well-defined classes of compounds was not predicted by this approach and required separate modeling.
Acknowledgment. Toxicity data acquisitions were supported in part by The University of Tennessee Center of Excellence in Livestock Disease and Human Health. Drs. Netzeva and Aptula were supported by the European Union IMAGETOX Research Training Network (HPRN-CT- 1999-00015). Gratitude is expressed to Mr. Glendon Sinks for his assistance with toxicity data analyses.
References (1) Green, S., Goldberg, A., and Zurlo J. (2001) TestSmart - high production volume chemicals: An approach to implementing alternatives into regulatory toxicology. Toxicol. Sci. 63, 6-14. (2) McKinney, J. D., Richard, A., Waller, C., Newman, M. C., and Gerberick, F. (2000) The practice of structure activity relationships (SAR) in toxicology. Toxicol. Sci. 56, 8-17. (3) Schultz, T. W., Sinks, G. D., and Bearden, A. P. (1998) QSAR in aquatic toxicology: A mechanism of action approach comparing toxic potency to Pimephales promelas, Tetrahymena pyriformis, and Vibrio fischeri. In Comparative QSAR (Devillers, J., Ed.) pp 52-109, Taylor and Francis, London. (4) Schultz, T. W. (1997) TETRATOX: The Tetrahymena pyriformis population growth impairment endpoint - A surrogate for fish lethality. Toxicol. Methods 7, 289-309. (5) Schultz, T. W., and Cronin, M. T.D. (2003) Essential and desirable characteristics of ecotoxicity QSARs. Environ. Toxicol. Chem. (in press). (6) Karcher, W., and Devillers, J. (1990) SAR and QSAR in environmental chemistry and toxicology: Scientific tool or wishful thinking? In Practical Applications of Quantitative StructureActivity Relationships (QSAR) in Environmental Chemistry and Toxicology (Karcher, W., and Devillers, J., Eds.) pp 1-12, Kluwer Academic, Dordrecht. (7) McFarland, J. W. (1970) On the parabolic relationship between drug potency and hydrophobicity. J. Med. Chem. 13, 1092-1196. (8) Mekenyan, O. G., and Veith, G. D. (1994) The electronic factor in QSAR: MO-parameters, competing interactions, reactivity and toxicity. SAR QSAR Environ. Res. 2, 129-143. (9) Karabunarliev, S., Mekenyan, O. G., Karcher, W., Russom, C. L., and Bradbury, S. P. (1996) Quantum-chemical descriptors for estimating the acute toxicity of electrophiles to the fathead minnow (Pimephales promelas): An analysis based on molecular mechanisms. Quant. Struct.-Act. Relation. 15, 302-310. (10) Bradbury, S. P. (1994) Predicting modes of toxic action from chemical structure: An overview. SAR QSAR Environ. Res. 2, 89-104. (11) Cronin, M. T. D., and Basketter, D. A. (1994) A multivariate QSAR analysis of a skin sensitization database. SAR QSAR Environ. Res. 2, 159-179. (12) SAS Institute Inc. (1989) SAS/STAT User’s Guide, version 6, 4th ed., Vol. 2, pp 846, North Carolina. (13) ClogP ver 1.0 for Windows (1995) BIOBYTE Corp., Claremont, CA. (14) TSAR ver 3.3 for Windows (2000) Accelrys, Oxford, England. (15) Stewart, J. J. P. (1993) MOPAC 93 Fujitsu Ltd., Windows 95/98/ NT/2k adaptation and MO indices - J. Kaneti (1988-94) MOQC. (16) MINITABTM ver 13.1 for Windows (2000) Minitab Inc., State College, Pennsylvania. (17) Mekenyan, O. G., Veith, G. D., Bradbury, S. P., and Russom, C. L. (1993) Structure-toxicity relationships for R,β-unsaturated alcohols in fish. Quant. Struct.-Act. Relat. 12, 132-136. (18) Hermens, J. L. M. (1990) Electrophiles and acute toxicity to fish. Environ. Health Perspect. 87, 219-225.
QSARs for Aliphatic Chemicals (19) Schultz, T. W., and Comeaux, J. L. (1996) Structure-toxicity relationships for aliphatic isothiocyanates to Tetrahymena pyriformis. Bull. Environ. Contam. Toxicol. 56, 638-642. (20) Sinks, G. D., Carver, T. A., and Schultz, T. W. (1998) Structuretoxicity relationships for aminoalcohols: A comparison with alkanols and alkanamines. SAR QSAR Environ. Res. 9, 217-228. (21) Bradbury, S. P., and Lipnick, R. L. (1990) Introduction: Structural properties for determining mechanisms of toxic action. Environ. Health Perspect. 87, 181-182. (22) Bearden, A. P., Sinks, G. D., and Schultz, T. W. (1999) Acclimation to sublethal exposure to a model nonpolar narcotic: Population growth kinetics and membrane lipid alterations in Tetrahymena pyriformis. Aquat. Toxicol. 46, 11-21. (23) Veith, G. D., Lipnick, R. L., and Russom, C. L. (1989) The toxicity of acetylenic alcohols to the fathead minnow, Pimephales promelas: Narcosis and proelectrophile activation. Xenobiotica 19, 555-565. (24) Veith, G. D., and Mekenyan, O. G. (1993) A QSAR approach for estimating the aquatic toxicity of soft electrophiles (QSAR for soft electrophiles). Quant. Struct.-Act. Relation. 12, 349-356. (25) Cronin, M. T. D., and Dearden, J. C. (1995) Review: QSAR in toxicology. 1. Prediction of aquatic toxicity. Quant. Struct.-Act. Relation. 14, 1-7. (26) Bradbury, S. P., Russom, C. L., Ankley, G. T., Schultz, T. W., and Walker, J. D. (2002) QSARs for predicting ecological effects of organic chemicals. Environ. Toxicol. Chem. (in press). (27) Akers, K. S., Sinks, G. D., and Schultz, T. W. (1999) Structuretoxicity relationships for selected halogenated aliphatic chemicals. Environ. Toxicol. Pharmacol. 7, 33-39. (28) DeWeese, A. D., and Schultz, T. W. (2001) Structure-activity relationships for aquatic toxicity to Tetrahymena: Halogensubstituted aliphatic esters. Environ. Toxicol. 16, 54-60. (29) Cronin, M. T. D., Sinks, G. D., and Schultz, T. W. (2001) Modelling of toxicity to the ciliate Tetrahymena pyriformis: the Aliphatic Carbonyl domain. In Forecasting the Environmental Fate and Effects of Chemicals (Rainbow, P. S., Hopkin, S. P., and Crane, M., Eds.) pp 113-124, John Wiley & Sons, Chichester, United Kingdom. (30) Dearden, J. C., and Bresnen, G. M. (1988) The measurement of partition coefficient. Quant. Struct.-Act. Relat. 7, 133-144.
Chem. Res. Toxicol., Vol. 15, No. 12, 2002 1609 (31) Deneer, J. W., Seinen, W., and Hermens, J. L. M. (1988) The acute toxicity of aldehydes to the guppy. Aquat. Toxicol. 12, 185-192. (32) Muccini, M., Layton, A. C., Sayler, G. S., and Schultz, T. W. (1999) Aquatic toxicities of halogenated benzoic acids to Tetrahymena pyriformis. Bull. Environ. Contam. Toxicol. 62, 616-622. (33) Cameron, I. L. (1973) Growth characteristics of Tetrahymena. In Biology of Tetrahymena (Elliott, A. M., Ed.) pp 199-226, Dowden, Hutchinson & Ross, Pennsylvania. (34) Seward, J. R., and Schultz, T. W. (1999) QSAR analyses of the toxicity of aliphatic carboxylic acids and salts to Tetrahymena pyriformis. SAR QSAR Environ. Res. 10, 557-567. (35) McKim, J. M., Bradbury, S. P., and Niemi, G. J. (1987) Fish acute toxicity syndromes and their use in the QSAR approach to hazard assessment. Environ. Health Perspect. 71, 171-186. (36) Austin, R. P., Davis, A. M., and Manners, C. N. (1995) Partitioning of ionizing molecules between aqueous buffers and phospholipid vesicles. J. Pharmacol. Sci. 84, 1180-1183. (37) Dearden, J. C., Bradburne, S. J. A., Cronin, M. T. D., and Solanki, P. (1988) In QSAR 88-Proceedings of the Third International Workshop on Quantitative Structure-Activity Relationships in Environmental Toxicology (Turner, J. E., England, M. W., Schultz, T. W., and Kwaak, N. J., Eds.) pp 43, USDOE, Oak Ridge, Tennessee. (38) Cronin, M. T. D., Dearden, J. C., Duffy, J. C., Edwards, R., Manga, N., Worth, A. P., and Worgan, A. D. P. (2002) The importance of hydrophobicity and electrophilicity descriptors in mechanisticallybased QSARs for toxicological endpoints. SAR QSAR Environ. Res. 13, 167-176. (39) Lipnick, R. L. (1991) Outliers, their origin and use in the classification of molecular mechanisms of toxicity. In QSAR in Environmental Toxicology - IV (Hermens, J. L. M, and Opperbuizen, A., Eds.) pp 131-153, Elsevier, Amsterdam, Netherlands. (40) Blaauboer, B. J., Barratt, M. D., and Houston, J. B. (1999) The integrated use of alternative methods in toxicological risk evaluation - ECVAM Integrated Testing Strategies task force report 1. ATLA 27, 229-237. (41) Egan, W. J., and Morgan, S. L. (1998) Outlier detection in multivariate analytical chemical data. Anal. Chem. 70, 2372-2379.
TX025589P
