Development of structure-activity relationship rules for predicting carcinogenic potential of chemicals

Development of structure-activity relationship rules for predicting carcinogenic potential of chemicals

Toxicology Letters 79 (1995) 219-228 Development of structure-activity relationship rules for pr#edicting carcinogenic potential of chemicals Yin-tak...

918KB Sizes 0 Downloads 21 Views

Toxicology Letters 79 (1995) 219-228

Development of structure-activity relationship rules for pr#edicting carcinogenic potential of chemicals Yin-tak Woo*, David Y. Lai, Mary F. Argus, Joseph C. Arcos Health and Environmental Review Division (7403), Ofi,ce of Pollution Prevention and Toxics, U.S. Environmental Protection Agency, 401 M Street SW, Washington, D.C. 20460, USA

Accepted 5 April 1995

Abstract

Since the inception of Section 5 (Premanufacturing/Premarketing Notification, PMN) of the Toxic Substances Control Act (TSCA), structure-activity relationship (SAR) analysis has been effectively used by U.S. Environmental Protection Agency’s (EPA) Structure Activity Team (SAT) in the assessment of potential carcinogenic hazard of new chemicals for which test data are not available. To capture, systematize and codify the Agency’s predictive expertise in order to make it more widely available to assessors outside the TSCA program, a cooperative project was initiated to develop a knowledge rule-based expert system to mimic the thinking and reasoning of the SAT. In this communication, we describe the overall structure of this expert system, discuss the scientific bases and principles of SAR analysis of chemical carcinogens used in the development of SAR knowledge rules, and delineate the major factors/rules useful for assessing the carcinogenic potential of fibers, polymers, metals/metalloids and several major classes of organic chemicals. An integrative approach using available short-term predictive tests and non-cancer toxicological data to supplement SAR analysis has also been described. Keywords:

Chemical carcinogen;

Structure-activity

relationship

analysis; Expert system; Hazard identification;

Risk

assessment

1. Introduction Abbreviations: EPA, U.S. Environmental Protection Agency; IARC, International Agency for Research on Cancer; NCI, U.S. National Cancer Institute; NTP, U.S. National Toxicology Program; PMN, Premanufacturing/Premarketing Notification; SAR, structure-activity relationships; TSCA, Toxic Substances Cont:rol Act. Disclaimer: The scientific views and ideas expressed in this communication are solely those of the authors and do not necessarily reflect those of the Agency. * Corresponding author. 0378-4274/95/$09.50 Elsevier Science Ireland Ltd. SSDI 0378-4274(95)011373-S

Section 5 of the Toxic Substances Control Act (TSCA) requires the U.S. Environmental Protection Agency (EPA) to assess potential toxic effects, including carcinogenicity, of new industrial chemicals within 90 days after the submission of a submitter’s notification of intent to begin manufacturing or importation. Since most of the submitted chemicals do not have test data. struc-

220

Y.-t. Woo et al. 1 Toxicology Letters 79 (1995) 219-228

ture-activity relationship (SAR) analysis has been routinely and effectively used by the EPA’s Structure Activity Team (SAT) in the assessment of carcinogenic and other toxic potential [l-3]. To capture, systematize and codify the Agency’s predictive expertise in order to make it more widely available to assessors outside the TSCA program, a cooperative project was initiated to develop a knowledge rule-based expert system to mimic the thinking and reasoning of the SAT. In this communication, we present a brief overview of this expert system, discuss the scientific bases and principles of SAR analysis of chemical carcinogens used in the development of SAR knowledge rules, and describe an integrative approach to use available short-term predictive tests and non-cancer toxicological data to supplement SAR analysis. 2. Overview of basic principles and approaches of SAR analysis Essentially, SAR analysis involves structural analogy of an untested chemical compound to related compounds, for which test data are available, combined with consideration of all relevant physicochemical and supportive data to arrive at a prediction of carcinogenic/toxic potential along with rationale for the prediction. The basic concepts and principles of SAR analysis have been described in detail in several recent reviews (e.g., [4-61). There are various approaches to SAR analysis, the SAT mainly uses a mechanism-based approach with semi-quantitative assessment of concern levels ranging from low to high. There are at least 5 critical factors that must be considered for any SAR analysis: (1) physicochemical; (2) molecular geometric; (3) electronic and steric; (4) metabolic; (5) mechanistic. Physicochemical properties (e.g., molecular weight, physical state, solubility, chemical reactivity) of a chemical compound determine its ability to reach target tissues or cells. Molecular size/shape/geometry and steric factors affect the ability to interact with target macromolecules as well as the chance to be metabolically activated or detoxified. Electronic factors may affect lifetime of reactive intermediate(s) through resonance stabilization as

well as interaction at the target macromolecules. Knowledge of metabolism of related compounds and assessment of the effect of substituents in the untested compound (e.g., blocking of detoxification, enhancement of activation pathways) are essential for SAR analysis. Mechanistic considerations (e.g., electrophilic, receptor-mediated, nonspecific cytotoxicity) are crucial for accurate SAR prediction. In addition, the projected mechanism of action may provide input to assessment of human relevance and affect the overall concern level. Since carcinogenesis is a multi-stage process, chemicals that affect multiple stages (i.e., initiation, promotion, progression) should be given greater concern. Most typical potent carcinogens: (1) demonstrate ability to reach target sites; (2) are activated or remain reactive near or at the target site; (3) have reasonable lifetime to allow interaction with target macromolecules; (4) display selective, specific and persistent interaction with target macromolecules; (5) affect multiple stages of carcinogenesis. Considerations of critical factors that impart carcinogenicity as well as those that abolish activity in individual chemical classes provide the basis for prediction of carcinogenic potential of untested compounds. 3. Benefits and goals of EPA’s cancer expert system, ‘OncoLogic’ An expert system, in simple terms, is a computerized system that mimics the thinking and reasoning of human experts. The benefits of an expert system include: (1) capture expertise; (2) reduce/eliminate error and inconsistency; (3) allow non-experts to arrive at expert judgment; (4) expedite decision making; (5) knowledge sharing. There are at least several different types of expert system, the expert system used in this project is a knowledge rule-based system. This type of system closely mirrors SAT’s thinking process and is ideal for the purpose. The goals of EPA’s expert system include: (1) to provide a system to bridge chemists and toxicologists for the most effective hazard evaluation; (2) to capture expertise of SAT in cancer hazard identification; (3) to streamline the TSCA Section

Y.-t. Woo et al. / Toxicology Letters 79 (1995) 219-228

5 evaluation process; (4) to provide guidance to industries on elements of concern for developing safer chemicals; (5) to stimulate research to fill knowledge gaps; (6) to provide a forum for reaching common understanding among various federal and state regulatory agencies in hazard identification/risk assessment of potential/known chemical carcinogens; (7) to provide a source of information to the public on rationale for identifying potential cancer hazard of chemicals. 4. Overview of the cancer expert system, ‘OncoLogic’ 4.1. Overall structure The overall structure of the cancer expert system is depicted in Fig. 1. It consists of a structural arm which evaluates carcinogenic potential by SAR analysis and a functional arm (see Section 4.5 below) which modifies the concern level from structural arm if non-cancer supportive data correlatable to carcinogenicity are available for the chemical in question. The structural arm, in turn, consists of 4 subsystems which evaluate fibers, polymers, metals, and organic chemicals. Each of these subsystems are built with specific knowledge rules that are unique for the subsystem to maximize accuracy of prediction. 4.2. Unique features The cancer expert system that we are developing has several unique features: (1) it is developed OVERALL STRUCTURE

OF THE CANCER EXPERT SYSTEM

\ Fig. 1. Overall structure of the cancer expert system.

*

221

using state-of-the-art expert system technology based on ‘knowledge rules’ which represent SAT’s formalized, codified and organized knowledge of SAR of carcinogens; (2) it has a flexible infrastructure which allows evaluation of virtually any type of chemical substances (fibers, polymers, metals, organic compounds). This is made possible because class-specific parameters are used in the evaluation of each class of chemical compounds to maximize accuracy; (3) the input includes not only the chemical structure but also all aspects (e.g., physicochemical properties and chemical stability, route of exposure, metabolic activation and mechanistic consideration, available genotoxicity and supportive data) critical to the evaluation of carcinogenic potential; (4) the output includes prediction/evaluation of carcinogenic potential along with underlying scientific rationale. Six concern levels (low, marginal, lowmoderate, moderate, high-moderate, high)’ are used to allow semi-quantitative ranking of relative hazard. 4.3. Knowledge rule development and sources of information The process for development of SAR rules for each structural class of chemical compounds include: (1) gather all available information on physicochemical properties, chemical carcinogenicity and related supportive data; (2) brainstorming to determine the critical structural features and secondary modifying factors that determine or contribute to carcinogenicity; (3) determine need for subclassification to optimize predictive capability; (4) narrow down to the top factors, assess usefulness in terms of user-friendli’ The criteria used to assign concern level include: (1) cancer bioassay dose-response data or comparison to reference compounds; (2) tumor pathology and multiplicity in species, sex and target; (3) mechanistic, pharmacokinetic, physicochemical consideration; (4) supportive evidence. Some representative carcinogens in the various concern levels are: (1) high: benzo[a]pyrene, N-nitroso-diethylamine, 1,2-dimethylhydrazine; (2) high-moderate: N-nitroso-piperidine, benzo[b]fluoranthene; (3) moderate: N-nitrosopyrrolidine, chloroethane; (4) lowmoderate: benz[a]anthracene, trimethyl phosphate; (5) marginal: benzo[e]pyrene, BHT, TPA; (6) low: pyrene, N-nitrosodioctylamine.

222

Y.-t. Woo et al. / Toxicology Letters 79 (1995) 219-228

ness and suitability for system development, and assign importance to each factor; (5) develop knowledge rules. A sample baseline knowledge rule (which establishes the concern level of a relatively simple compound) and a sample mobile knowledge rule (which establishes the modifying effect of a certain substituent on the baseline concern level) are illustrated below: Baseline rule # xx for AA-l: IF - A 6-membered homocyclic aromatic compound with 1 amino group or amine-generating group bears on the ring: 1 or 2 methyl group(s) at least one of which is o&o to the amino or amine-generating group of the ring THEN - it is very likely (0.80) that the compound is carcinogenic. Concern level: high-moderate Mobile rule # yy for AA-l: IF - Into a 6-membered homocyclic aromatic compound containing 1 amino or amine-generating group and 1 ring methyl/methoxy/ethyl/ethoxy group, 1 or 2 ring chloro/bromo substituent(s) is(are) introduced additionally THEN - the concern level will be increased by 1 level or will be high-moderate, whichever is lower. All these knowledge rules for each class or subclass of chemical compounds are interwoven into intricate decision trees. All the decision trees, in turn, are interconnected through an interface manager software which guides the user to the appropriate decision tree through queries and structure/data input. Whenever necessary, specialized software is designed to handle specific complicated calculations (e.g., incumbrance area of polynuclear aromatic hydrocarbons) or automatic graphic identification of specific molecular structural features (e.g., longest resonance-stabilized conjugated chain in aromatic systems). The major sources of information used in developing knowledge rules for the current cancer expert system include: (1) the ‘Chemical Induction of Cancer’ monograph series [7- 111;(2) International Agency for Research on Cancer (IARC) monograph series; (3) National Cancer Institute (NCI)/ National Toxicology Program (NTP) technical reports; (4) the U.S. Public Health Services Publication 149 series, ‘Survey of Compounds which have been Tested for Carcinogenic Activity’; (5)

non-classified EPA submission data from various program offices. Whenever necessary, literature searches/reviews and consultation with researchers are conducted to ensure that the information is up to date. 4.4. The structural arm: rules and critical factors/ features The structural arm is designed to evaluate carcinogenic potential of untested chemicals based on SAR analysis and to assess relative carcinogenic hazard of tested compounds by assigning a concern level in comparison to other tested carcinogens. It consists of 4 subsystems as described below. (1) The Fiber Subsystem is designed for evaluation of fibrous substances. It is well documented [l l] that the most critical factors that contribute to carcinogenicity of fibers are: physical dimensions (i.e., fiber diameter and aspect ratio) and physicochemical properties (e.g., surface charge, flexibility, durability). Beyond these, the chemical composition of the fibers does not appear to be a significant contributing factor. Using the data of Stanton et al. [12] as a guide, we have developed a ‘grid’ that gives the initial concern level of the fiber (ranging from low to high-moderate) based on the information of fiber size that the user enters into the computer. The initial concern level is then modified according to information on physicochemical properties and relevant manufacturing/processing/ use information to arrive at the final concern level. (2) The Polymer Subsystem is designed for evaluation of high-molecular-weight polymeric compounds which are often (in some cases, erroneously) considered as ‘safe’ but can be hazardous under certain conditions. The factors used for evaluation of polymers include: (1) extent of presence of low-molecular-weight species; (2) extent and nature of reactive functional group(s) present; (3) solubility and ability to swell; (4) special features such as polysulfation; (5) exposure scenario (e.g., possibility of inhalational exposure to the degree of lung overloading); (6) possibility of breakdown and ensuing products. Menus graphically displaying reactive functional groups and polymeric linkages are included to maximize user-friendliness.

Y.-t. Woo et al. / Toxicology Letters 79 (1995) 219-228

(3) The Metal/hAetalloid Subsystem is designed for evaluation of inorganic and organic chemicals that contain metals or metalloids. The elements of concern considereld include: (1) the nature of the metal/metalloid; (2) the type of chemical bonding (e.g., ionic, coordination, organometallic); (3) dissociability and solubility of the compound; (4) oxidation state of the metal or metalloid; (5) physical state (e.g., if solid, crystalline or amorphous); (6) exposure scenario (e.g., possibility of inhalation exposure); (7) possible breakdown products (e.g., organic ligand). By taking these factors into consideration, 2 compounds containing the same metal may have vastly different concern levels (e.g., high for inhalation exposure to a sparingly soluble, crystalline Cr(V1) compound vs. marginal for dermal exposure to an insoluble, amorphous Cr(II1) compound). (4) The Orgamc Subsystem contains over 30 chemical classes (see Table 1). Each of these classes has its own class-specific knowledge rules which maximize predictive capability. When necessary, a chemical class may be further subdivided into subclasses for the development of subclassspecific knowledge rules. To exemplify the process for rule development and scientific thinking involved, the procedure leading to the development of knowledge rules for 3 major structural classes of chemicals is briefly described below. Polynuclear aromatic hydrocarbons (PAHs) represent one of the most extensively studied classes of chemica.1 carcinogens [7,13,14]. The key factors that contribute to carcinogenic activity include: (1) a favorable molecular size and shape; (2) capability to generate reactive intermediates (e.g., carbonium ion, free radical) after metabolic activation; (3) capability to provide resonance stabilization of reactive intermediates to allow sufficient lifetime to reach target macromolecules. For the PAHs, incumbrance area*, an effective method to measure molecular size and shape, has

2The incumbrance area of a polynuclear compound is defined as the area (in A’) of the smallest rectangular envelope that can contain the entire planar molecule of the polynuclear compound drawn proportionally to molecular dimensions. The ratio of the length/width of the rectangular envelope is called the incumbrance aspect ratio (see Ref. [7] for detail).

223

Table 1 Major chemical classes covered in the organic subsystem Polynuclear aromatic hydrocarbons (homocyclic and heterocyclic) Aromatic amine compounds Arylazo compounds N-Nitroso compounds Hydrazo compounds, aliphatic azo and azoxy compounds and triazenes Halogenated linear alkanes and alkenes Halogenated cycloalkanes, cycloalkenes and aromatic hydrocarbons Halogenated phenoxy acids, aromatic ethers, dibenzofurans and dibenzo-p-dioxins Carbamates, thiocarbamates and substituted urea compounds Phenols and phenolic compounds Nitrogen and sulfur mustards Haloethers and halothioethers Epoxides and aziridines Lactones and sultones Alkyl sulfates and alkyl alkanesulfonates Aldehydes Reactive ketones and sulfones Nitroalkanes and nitroalkenes Thiocarbonyl compounds Alkenyl benzene compounds Organophosphorus compounds Peroxides and peroxy compounds C-Nitroso compounds and oximes Coumarins and furocoumarins Acrylates and methacrylates Acylating agents

been found to be highly useful as an initial indicator of potential activity [7]. Virtually all PAHs with incumbrance area exceeding 185 A* or aspect ratio exceeding 2.1 have been consistently shown to be inactive. A computer program has been written to handle the laborious calculation of incumbrance area. In evaluating each PAH, all possible metabolic activating pathways (e.g., bayregion diolepoxide formation, one-electron oxidation, biomethylation) are taken into account. The following structural features have been found to impart carcinogenic activity: (1) presence of unsubstituted bay-region/pseudo bay-region/gulf region benzo ring; (2) unoccupied peri position adjacent to the bay-region benzo ring (e.g., 5-position of benz[u]anthracene), referred to as ‘ -P effect’ if occupied; (3) substitution at the peri position (inside the bay-region) of inner naphtho

224

Y.-t. Woo et al. J Toxicology Letters 79 (1995) 219-228

moiety of the bay-region benzo ring with methyl group (e.g., 5-position of chrysene), referred to as ‘ + B effect’; (4) substitution at the L-region with methyl group(s); (5) lack of bulky alkyl or highly hydrophilic substituents, particularly at critical positions or regions (e.g., L-, - P, + B, bay-region). They have been incorporated into knowledge rules and decision trees for SAR analysis. A computer program capable of graphically locating these sensitive regions and positions has been compteted. Aromatic amines represent another major class of carcinogenic chemicals [8,15,16]. The critical factors include: (1) molecular size, shape and planarity; (2) number and position (e.g., relative to the longest resonance pathway of the aromatic system) of amine group or amine-generating group; (3) possibility of metabolic activation; (4) presence or absence of substituents, particularly those at the amino nitrogen and those flanking the amino group. Depending on the number of homocyclic or heterocyclic aromatic ring(s) and the number of amine or amine-generating groups, this Component has been subdivided into 9 subcomponents to maximize predictive capability. Fig. 2 shows a sample printout of an aromatic amine with 2 aromatic rings linked by an intercyclic bond. The example illustrates the rationale of arriving at a final concern level of low-moderate. If this compound is devoid of any substituents on the ring and the amino nitrogen, it would be given a high concern level because the molecule has a favorable planar structure with the amine and amine-generating groups occupying the most favorable 4,4’-positions in terms of resonance stabilization of putative metabolically activated nitrenium ions. The substituents on the ring and amine nitrogen together with the intercyclic linkage have a combined effect of reducing to low-moderate as illustrated. In addition, it has been pointed out that the intercyclic linkage may be hydrolyzed in acidic condition to give degradation products which should be evaluated separately in an appropriate Organic Component. The Aromatic Amine Component, the first component of the Organic Subsystem, was used to predict the carcinogenic potential of 8 aromatic amines before data on U.S. NTP bioassay were available to

us. Its predictive performance is discussed under Section 4.6 below. The N-nitroso compounds component consists of N-nitrosamines (which is subdivided into

Fig. 2. Sample printout of an assessment and justification report for the potential cancer hazard of a chemical compound using the cancer expert system.

225

Y.-t. Woo et al. / Toxicology Letters 79 (1995) 219-228

Table 2 A tentative list of endpoints for the functional arm and their putative major contribution to carcinogenesis Tox./Pharmacol.

Endpoint

A. Oncogene activation B. Cell transformation C. Mutagenicity and related (1) Gene/point mutation (2) Chromosome aberration (3) Aneuploidy D. DNA/RNA binding E. Electrophilicity F. Immunosuppressant H. Hormonal imbalanc:e (e.g., estrogenic, thyroid, beta-adrenergic) I. Antineoplastic agent J. Peroxisome proliferator K. Target tissue damage L. Cell proliferation N. Oxy/free radical producer 0. DNA hypomethylation P. GapJuncIntCom inhibitor Q. Mitochondrial inhibitor R. Microsomal enzyme: inducer S. Purine/pyrimidine analog T. Mitogens U. Growth factor inducers V. Surface active agent W. Strong chelator X. Cell membrane disruptor

Init.

Promo.

Progr.

X

X

X

x?

x?

x?

X

x?

X

x?

Unknown

X X x?

X

X X

x?

x? x?

x? x?

x?

X X X

x?

X

X X X X

nitroand cycloaliphatic dialkylnitrosamines samines) and N-nitrosamides (which is subdivided into N-nitrosoureas, N-nitrosoguanidines, N-nitrosocarbamates, N-nitrosocyanamides, N-nitrosobenzamides and N-nitrosocarboxylamides). For the N-nitrosamines, the critical factors are the type of alkyl or cycloalkyl group and the possibility of metabolic activation (mostly by ar-hydroxylation but in some specific cases by p-, y- or w-hydroxylation). For the N-nitrosamides, the critical factors are the type of alkyl group and the possibility of alkali-, thiol- or esterase-catalyzed hydrolysis. In both cases, substitution with highly bulky or hydrophilic groups almost invariably decreases or even abolishes activity particularly at sensitive positions such as a-carbon [9,17]. Owing to the wealth of data available, the N-nitroso component evaluates both untested compounds by SAR analysis and tested compounds by giving a relative hazard concern level.

X X X

X X X X X

4.5. The functional arm This arm is currently under development and will be used to modify and/or substantiate the concern level from the structural arm if non-cancer toxicological data known to correlate with carcinogenicity are available for the chemical being evaluated. The arm will contain organized information on structural< class-specific predictive capability of various short-term tests and noncancer toxicological endpoints. A number of such endpoints are listed in Table 2 along with their possible contribution to the various stages in the multi-stage carcinogenesis process. If data on only one or a few tests/endpoints are available, they will be used to modify the concern level from structural arm only if the test(s) was(were) evaluated and considered to be suitable and adequate for the test compound and the test(s) have been found to have high predictive power for compounds of the same structural class as the chemi-

226

Y.-t. Woo et al. 1 Toxicology

cal in question. If data on multiple tests/endpoints are available, an integrative approach (see Fig. 3) may be used to arrive at a concern level and then compared to the concern level from the structural arm. 4.6. Validation and verification issues For any predictive system, prospective validation is probably the only foolproof method of testing the accuracy of prediction. In a recent workshop on comparison of various cancer predictive systems, it appears that the system that utilizes human expertise to consider both the structural alerts and biological activities, fared best in that particular predictive exercise [18]. The current OncoLogic system basically uses a similar approach but is formalized and codified so that the presence of human experts is not required. The OncoLogic system was not included in the predictive exercise mentioned above because it was still at the early stage of development at the

’ [pgEiJs&] ’

Letters

79 (1995) 219-228

time of commencement of the predictive exercise. Nevertheless, predictions on 8 aromatic amines were made before the bioassay results were available. Table 3 shows a comparison of predictions vs. actual results. Of these, the clearly potent carcinogens (compounds # 40,43) and the clearly negative compound ( # 24) were given concern levels of high-moderate and low, respectively. Compound # 32, a one-species positive carcinogen, was given a rating of moderate. The 2 equivocal carcinogens, compounds # 42 and # 33, were rated marginal and low-moderate, respectively. Compound # 26, a very weak carcinogen (the tumor data on this compound were so weak that the NTP at one time considered it equivocal), was also given a marginal rating. Compound # 33, which was rated low-moderate, was found to be negative in the mouse but has yet to be tested in the rat. Overall, the semi-quantitative concern level predictions closely parallel the actual results. Beyond the specific examples cited above, the OncoLogic system has also been continuously verified and validated as new bioassay data become available. A Peer Review Pane13, consisting of the leading experts in cancer SAR analysis and artificial intelligence, conducted a thorough evaluation of the OncoLogic system in September 1992 and confirmed its scientific validity. External experts have also been invited to provide input useful for rule development and evaluate specific Organic Components on a continuous basis. Additional prospective validation studies are currently being planned. Acknowledgement

Fig. 3. An integrative approach of combining complementary short-term tests as supportive evidence for carcinogenicity. (The 3 circles represent 3 categories of short-term tests indicative of activity in the 3 stages of carcinogenesis: initiation (init.), promotion (promo.) and progression (progr.). The concern level of a chemical compound will depend on the number of categories shown to be positive. For example, chemical compounds which have been shown to yield clearly and consistently positive results in all 3 categories of short-term tests will be given a HM/H concern.)

The authors thank Dr. Marilyn Arnott of LogiChem. Inc., the Cooperative Agreement recipient, for providing the justification report used in this manuscript. Thanks are also due to Dr. Ernest Falke for providing administrative support and for preparing some of the illustrations. The 3The members of the Peer Review Panel include: Drs. James Miller, Subhash Basak, Ercole Cavalieri, Arthur Furst, William Lijinsky, Robert Lipnick, Danuta Malejka-Giganti, and Joe Stuart.

Y.-t. Woo et al. / Toxicology

Letters

221

79 (1995) 219-228

Table 3 OncoLogic@ prediction vs. NTP bioassay results aromatic amines and related compounds

24 42 26 9 33 32 40 43

Bioassay results

Chemical

#

4,4’-Diamino-2,2’-stilbene disulfonic acid p-Nitroaniline p-Nitrobenzoic acid p-Nitrophenol 4-Hydroxyacetanilide 2,4-Diaminophenol dihydrochloride 3,3’-Dimethylbenzidine o-Nitroanisole

OncoLogica evaluation

Rat

Mouse

‘Call’

N/N NT N/S NT N/E N/N C/C C/C

N/N E/N N/N N/N N/N S/N NT C/C

_ Eq + Eq + + +

L mar mar LM LM M HM HM

‘OncoLogic’ predictions of these compounds were made approximately 2 years before the publication of NTP bioassay results. The chemical # ‘s correspond to the numbers used in the NTP predictive exercise [IS]. The bioassay results (males/females) in rats and mice are summarized using the symbols below. ‘Call’ refers to overall rating of carcinogenicity considering all available rodent bioassay results. C, clear evidence of carcinogenicity; S, some evidence of carcinogenicity; E, equivocal evidence of carcinogenicity; N, no evidence of carcinogenicity; NT, not tested; + , at least one test = C or S; Eq, no C or S, and E must appear at least once; - , no C, S or E.

helpful discussion of Dr. Kirk Kitchin greatly appreciated.

is also

References

VI Arcos, J.C. (1983) ‘Comparative requirements for premarketing/premanufacture notifications in the EC countries and the USA with special reference to risk assessment in the framework of the U.S. Toxic Substances Control Act (TSCA). J. Am. Coil. Toxicol. 2, l31- 145. PI DiCarlo, F.J., Bickart, P. and Auer, C.M. (1986) Structure-metabolism relationships for the prediction of health hazards by the Environmental Protection Agency. Drug Metab. Rev. 17, 17-184. [31 Auer, C.M. and Gould, D.H. (1987) Carcinogenicity assessment and the role of structure activity relationship (SAR) analysis under TSCA Section 5. Environ. Carcinon. Rev. (Pt. C of J. Environ. Sci. Health) CS, 29-71. (41 Woo, Y.-t., Arcos, J.C. and Lai, D.Y. (1985) Structural and functional criteria for suspecting chemical compounds of carcinogenic activity: state-of-the-art of predictive formalism. In: H.A. Milman and E.K. Weisburger (Eds.), Handbook of Carcinogen Testing, Chapter I, Noyes Publication, Park Ridge, N.J., pp. 1-25. 151 Woo, Y.-t. and Amos, J.C. (1989) Role of structure-activity relationship analysis in evaluation of pesticides for potential carcinogenicity. In: N.N. Ragsdale and R.E. Menzel (Eds.), Carcinogenicity and Pesticides, ACS Symposium Series No. 414, American Chemical Society, Washington, D.C., pp. 175-200. WI Richard, A.M., Ftabinowitz, J.R. and Waters, M.D. (1989) Strategies for the use of computation SAR methods in assessing genotoxicity. Mutat. Res. 221, ISI- 196.

[7] Arcos, J.C. and Argus, M.F. (1974) Chemical Induction of Cancer: Structural Bases and Biologic Mechanisms. Vol. IIA, Polynuclear Aromatic Hydrocarbons. Academic Press, New York. PI Arcos, J.C. and Argus, M.F. (1974) Chemical Induction of Cancer: Structural Bases and Biologic Mechanisms. Vol. IIB, Aromatic Amines and Related Compounds. Academic Press, New York. 191Arcos, J.C., Woo, Y.-t. and Argus, M.F. (with collaboration of Lai, D.Y.) (1981) Chemical Induction of Cancer: Structural Bases and Biologic Mechanisms. Vol. IIIA, Aliphatic Carcinogens. Academic Press, New York. WI Woo, Y.-t., Lai, D.Y., Arcos, J.C. and Argus, M.F. (1985) Chemical Induction of Cancer: Structural Bases and Biologic Mechanisms. Vol. IIIB, Aliphatic and Polyhalogenated Carcinogens. Academic Press, Orlando, Fla. D]l Woo, Y.-t., Lai, D.Y., Arcos, J.C. and Argus, M.F. (1988) Chemical Induction of Cancer: Structural Bases and Biologic Mechanisms. Vol. IIIC, Natural, Metal, Fiber and Macromolecular Carcinogens. Academic Press, San Diego, Calif. 1121Stanton, M.F., Layard, M., Tegeris, A., Miller, E., May, M., Morgan, E. and Smith, A. (1981) Relation of particle dimension to carcinogenicity in amphibole asbestos and other fibrous materials. J. Nat]. Cancer Inst. 67, 965-976. 1131Yang, SK. and Silverman, B.D. (Eds.) (1988) Polycyclic Aromatic Hydrocarbon Carcinogenesis: Structure Activity Relationships. Vols. I and II. CRC Press, Boca Raton, Fla. 1’41 Richard, A.M. and Woo, Y.-t. (1990) A CASE-SAR analysis of polycyclic aromatic hydrocarbon carcinogenicity. Mutat. Res. 242, 285-303. [I51 Garner, R.C., Martin, C.N. and Clayson, D.B. (1984) Carcinogenic Aromatic Amines and Related Compounds.

228

Y.-t. Woo et al. / Toxicology Letters 79 (1995) 219-228

In: C.E. Searle (Ed.), Chemical Carcinogens, 2nd Edn., ACS Monograph 182, American Chemical Society, Washington, D.C. [16] Lai, D.Y., Woo, Y.-t., Argus, M.F. and Arcos, J.C. (1994) Structural Elements of Consideration for Assessing the Carcinogenic Hazard/Risk of New and Untested Aromatic Amines. Toxicologist 14, 134, Abstract #453.

[l7] Lijinsky, W. (1992) Chemistry and Biology of N-Nitroso Compounds. Cambridge University Press, New York. [18] Wachsman, J.T., Bristol, D.W., Spalding, J., Shelby, M. and Tennant, R.W. (1993) Predicting chemical carcinogenesis in rodents. Environ. Health Perspcct. 101, 44& 445.