Applications of mechanistic data in risk assessment: the past, present, and future

Investigating the mechanisms of action of thyroid disruptors: A multimodal approach that integrates in vitro and metabolomic analysis

The thyroid gland, a vital component of the endocrine system, plays a pivotal role in regulating metabolic processes, growth, and development. To better characterize thyroid system disrupting chemicals (TSDC), we followed the next-generation risk assessment approach, which further considers the mechanistic profile of xenobiotics. We combined targeted in vitro testing with untargeted metabolomics. Four known TSDC, propyl-thiouracil (PTU), sodium perchlorate, triclosan, and 5-pregnen-3β-ol-20-one-16α‑carbonitrile (PCN) were investigated using rat in vitro models, including primary hepatocytes, PCCL3 cells, thyroid microsomes, and three-dimensional thyroid follicles. We confirmed each compound's mode of action, PTU inhibited thyroperoxidase activity and thyroid hormones secretion in thyroid cells model, sodium perchlorate induced a NIS-mediated iodide uptake decrease as triclosan to a lesser extent, and PCN activated expression and activity of hepatic enzymes (CYPs and UGTs) involved in thyroid hormones metabolism. In parallel, we characterized intracellular metabolites of interest. We identified disrupted basal metabolic pathways, but also metabolites directly linked to the compound's mode of action as tyrosine derivates for sodium perchlorate and triclosan, bile acids involved in beta-oxidation, and precursors of cytochrome P450 synthesis for PCN. This pilot study has provided metabolomic fingerprinting of dedicated TSDC exposures, which could be used to screen and differentiate specific modes of action.

Data-Driven Quantitative Structure-Activity Relationship Modeling for Human Carcinogenicity by Chronic Oral Exposure

Traditional methodologies for assessing chemical toxicity are expensive and time-consuming. Computational modeling approaches have emerged as low-cost alternatives, especially those used to develop quantitative structure-activity relationship (QSAR) models. However, conventional QSAR models have limited training data, leading to low predictivity for new compounds. We developed a data-driven modeling approach for constructing carcinogenicity-related models and used these models to identify potential new human carcinogens. To this goal, we used a probe carcinogen dataset from the US Environmental Protection Agency's Integrated Risk Information System (IRIS) to identify relevant PubChem bioassays. Responses of 25 PubChem assays were significantly relevant to carcinogenicity. Eight assays inferred carcinogenicity predictivity and were selected for QSAR model training. Using 5 machine learning algorithms and 3 types of chemical fingerprints, 15 QSAR models were developed for each PubChem assay dataset. These models showed acceptable predictivity during 5-fold cross-validation (average CCR = 0.71). Using our QSAR models, we can correctly predict and rank 342 IRIS compounds' carcinogenic potentials (PPV = 0.72). The models predicted potential new carcinogens, which were validated by a literature search. This study portends an automated technique that can be applied to prioritize potential toxicants using validated QSAR models based on extensive training sets from public data resources.

Gut microbiota-mediated pesticide toxicity in humans: Methodological issues and challenges in the risk assessment of pesticides

Many in vivo and in vitro studies have shown that pesticides can disrupt the functioning of gut microbiota (GM), which can lead to many diseases in humans. While the tests developed by the Organization of Economic Cooperation and Development (OECD) are expected to capture most apical effects resulting from GM disruptions, exclusion of GM in the risk assessment might mischaracterize hazards or overestimate/underestimate risks, especially when extrapolating results from one species to another species or population with a substantially different GM. On the other hand, direct assessment of GM-mediated effects may face challenges in identifying hazards, since not all GM perturbations will lead to human adverse effects. In this regard, reliable and validated biomarkers for common GM-mediated adverse effects may be very useful in the identification of GM-mediated pesticide toxicity. Nevertheless, proving causality of GM-mediated effects will need modifications of Bradford Hill criteria as well as Koch's postulates, which are more suitable for the "one-pathogen" paradigm. Furthermore, risk assessment of GM-mediated effects may require pesticide toxicokinetics along the gut, possibly through modeling, and the establishment of the involvement of GM in the mechanism of action (MOA) of the pesticide. Risk assessment of GM mediated effects also requires the standardization of experimental approaches as well as the establishment of microbial reference communities, since variations exist among GM in human populations.

Quantitative Interpretation of Genetic Toxicity Dose-Response Data for Risk Assessment and Regulatory Decision-Making: Current Status and Emerging Priorities

The screen-and-bin approach for interpretation of genotoxicity data is predicated on three false assumptions: that genotoxicants are rare, that genotoxicity dose-response functions do not contain a low-dose region mechanistically characterized by zero-order kinetics, and that genotoxicity is not a bona fide toxicological endpoint. Consequently, there is a need to develop and implement quantitative methods to interpret genotoxicity dose-response data for risk assessment and regulatory decision-making. Standardized methods to analyze dose-response data, and determine point-of-departure (PoD) metrics, have been established; the most robust PoD is the benchmark dose (BMD). However, there are no standards for regulatory interpretation of mutagenicity BMDs. Although 5-10% is often used as a critical effect size (CES) for BMD determination, values for genotoxicity endpoints have not been established. The use of BMDs to determine health-based guidance values (HBGVs) requires assessment factors (AFs) to account for interspecies differences and variability in human sensitivity. Default AFs used for other endpoints may not be appropriate for interpretation of in vivo mutagenicity BMDs. Analyses of published dose-response data showing the effects of compensatory pathway deficiency indicate that AFs for sensitivity differences should be in the range of 2-20. Additional analyses indicate that the AF to compensate for short treatment durations should be in the range of 5-15. Future work should use available data to empirically determine endpoint-specific CES values; similarly, to determine AF values for BMD adjustment. Future work should also evaluate the ability to use in vitro dose-response data for risk assessment, and the utility of probabilistic methods for determination of mutagenicity HBGVs. Environ. Mol. Mutagen. 61:66-83, 2020. © 2019 Her Majesty the Queen in Right of Canada.

Aquatic ecotoxicological models and their applicability in Arctic regions

Dose-response modeling is one of the most important steps of ecological risk assessment. It requires concentration-effects relationships for the species under consideration. There are very limited studies and experimental data available for the Arctic aquatic species. Lack of toxicity data hinders obtaining dose-response relationships for lethal (LC50 values), sub-lethal and carcinogenic effects. Gaps in toxicity data could be filled using a variety of in-silico ecotoxicological methods. This paper reviews the suitability of such methods for the Arctic scenario. Mechanistic approaches like toxicokinetic and toxicodynamic analysis are found to be better suited for interspecies extrapolation than statistical methods, such as Quantitative Structure-Activity Relationships/Quantitative Structure Activity-Activity Relationship, ICE, and other empirical models, such as Haber's law and Ostwald's equation. A novel approach is proposed where the effects of the toxicant exposure are quantified based on the probability of cellular damage and metabolites interactions. This approach recommends modeling cellular damage using a toxicodynamic model and physiology or metabolites interactions using a toxicokinetic model. Together, these models provide more reliable estimates of toxicity in the Arctic aquatic species, which will assist in conducting ecological risk assessment of Arctic environment.

In silico toxicology: computational methods for the prediction of chemical toxicity

Determining the toxicity of chemicals is necessary to identify their harmful effects on humans, animals, plants, or the environment. It is also one of the main steps in drug design. Animal models have been used for a long time for toxicity testing. However, in vivo animal tests are constrained by time, ethical considerations, and financial burden. Therefore, computational methods for estimating the toxicity of chemicals are considered useful. In silico toxicology is one type of toxicity assessment that uses computational methods to analyze, simulate, visualize, or predict the toxicity of chemicals. In silico toxicology aims to complement existing toxicity tests to predict toxicity, prioritize chemicals, guide toxicity tests, and minimize late-stage failures in drugs design. There are various methods for generating models to predict toxicity endpoints. We provide a comprehensive overview, explain, and compare the strengths and weaknesses of the existing modeling methods and algorithms for toxicity prediction with a particular (but not exclusive) emphasis on computational tools that can implement these methods and refer to expert systems that deploy the prediction models. Finally, we briefly review a number of new research directions in in silico toxicology and provide recommendations for designing in silico models. WIREs Comput Mol Sci 2016, 6:147-172. doi: 10.1002/wcms.1240 For further resources related to this article, please visit the WIREs website.

Background, approaches and recent trends for setting health-based occupational exposure limits: a minireview

The setting of occupational exposure limits (OELs) are founded in occupational medicine and the predictive toxicological testing, resulting in exposure-response relationships. For compounds where a No-Observed-Adverse-Effect-Level (NOAEL) can be established, health-based OELs are set by dividing the NOAEL of the critical effect by an overall uncertainty factor. Possibly, the approach may also be used for carcinogens if the mechanism is epigenetic or the genetic effect is secondary to effect from reactions with proteins such as topoisomerase inhibitors, and mitotic and meiotic spindle poisons. Additionally, the NOAEL approach may also be used for compounds with weak genotoxic effect, playing no or only a minor role in the development of tumours. No health-based OEL can be set for direct-acting genotoxic compounds where the life-time risks may be estimated from the low-dose linear non-threshold extrapolation, allowing a politically based exposure level to be set. OELs are set by several agencies in the US and Europe, but also in-house in major chemical and pharmaceutical companies. The benchmark dose approach may in the future be used where it has advantage over the NOAEL approach. Also, more attention should be devoted to sensitive groups, toxicological mechanisms and interactions as most workplace exposures are mixtures.

Mode of action frameworks: a critical analysis

Mode of action (MOA) information is increasingly being applied in human health risk assessment. The MOA can inform issues such as the relevance of observed effects in laboratory animals to humans, and the variability of response within the human population. Several collaborative groups have developed frameworks for analyzing and utilizing MOA information in human health risk assessment of environmental carcinogens and toxins, including the International Programme on Chemical Safety, International Life Sciences Institute, and U.S. Environmental Protection Agency. With the goal of identifying gaps and opportunities for progress, we critically evaluate several of these MOA frameworks. Despite continued improvement in incorporating biological data in human health risk assessment, several notable challenges remain. These include articulation of the significant role of scientific judgment in establishing an MOA and its relevance to humans. In addition, binary (yes/no) decisions can inappropriately exclude consideration of data that may nonetheless be informative to the overall assessment of risk. Indeed, the frameworks lack a broad consideration of known causes of human disease and the potential for chemical effects to act additively with these as well as endogenous background processes. No integrated analysis of the impact of multiple MOAs over the same dose range, or of varying MOAs at different life stages, is included. Separate consideration of each MOA and outcome limits understanding of how multiple metabolites, modes, and toxicity pathways contribute to the toxicological profile of the chemical. An extension of the analyses across outcomes with common modes is also needed.

Deriving uncertainty factors for threshold chemical contaminants in drinking water

Uncertainty factors are used in the development of drinking-water guidelines to account for uncertainties in the database, including extrapolations of toxicity from animal studies and variability within humans, which result in some uncertainty about risk. The application of uncertainty factors is entrenched in toxicological risk assessment worldwide, but is not applied consistently. This report, prepared in collaboration with Health Canada, provides an assessment of the derivation of the uncertainty factor assumptions used in developing drinking-water quality guidelines for chemical contaminants. Assumptions used by Health Canada in the development of guidelines were compared to several other major regulatory jurisdictions. This assessment has revealed that uncertainty factor assumptions have been substantially influenced by historical practice. While the application of specific uncertainty factors appears to be well entrenched in regulatory practice, a well-documented and disciplined basis for the selection of these factors was not apparent in any of the literature supporting the default assumptions of Canada, the United States, Australia, or the World Health Organization. While there is a basic scheme used in most cases in developing drinking-water quality guidelines for nonthreshold contaminants by the jurisdictions included in this report, additional factors are sometimes included to account for other areas of uncertainty. These factors may include extrapolating subchronic data to anticipated chronic exposure, or use of a LOAEL instead of a NOAEL. The default value attributed to each uncertainty factor is generally a factor of 3 or 10; however, again, no comprehensive guidance to develop and apply these additional uncertainty factors was evident from the literature reviewed. A decision tree has been developed to provide guidance for selection of appropriate uncertainty factors, to account for the range of uncertainty encountered in the risk assessment process. Recent development of a series of "decision trees" by WHO to derive chemical specific adjustment factors for inter- and intraspecies variability may present an opportunity for a more systematic approach for the identification of evidence-based uncertainty factors.

Analysis of reproductive toxicity and classification of glufosinate-ammonium

CONCLUSION REGARDING CLASSIFICATION OF GLUFOSINATE-AMMONIUM: Science Partners' Evaluation Group (Evaluation Group) has conducted an independent analysis of the herbicide glufosinate-ammonium (GA) relative to its potential to cause reproductive toxicity in humans. Further, the Evaluation Group has evaluated the implementation of Annex 6 of Commission Directive 2001/59/EC (28th ATP of Council Directive 67/548/EEC) and Council Directive 91/414/EEC, with respect to classification of chemicals posing potential reproductive hazards. After consideration of all information available to us relevant to the potential of glufosinate-ammonium (GA) to cause reproductive toxicity, the Science Partners Evaluation Group concludes that no classification of GA is justified. The following form the basis of this conclusion. There are no human data to suggest that GA causes reproductive toxicity in women or in their conceptus. The issue concerning possible reproductive hazard to humans is raised solely on the basis of positive animal test results that show GA to cause preimplantation or implantation losses in rats. SPECIFICALLY: a. Daily treatment with GA had no detectable effect on the earliest stages of the reproductive sequence including gametogenesis, ovulation, mating and conception; b. Treatment with GA interfered with rat gestation before and at the stage when the conceptus implants into the uterus. This effect occurred at doses of 360 ppm in the feed (corresponding to daily doses of 27.8 mg/kg bw) and above; and c. After implantation, no further effect of GA on prenatal and post-natal development was recognized. Previous concerns that GA might be toxic to embryonic stages after implantation were not supported by the data. Abortions and stillbirth seen were associated with, and regarded as secondary to, maternal toxicity. There was no evidence suggesting the induction of malformations in the offspring. The mechanism underlying this adverse effect in experimental laboratory animals is identified-inhibition of glutamine synthetase. Glutamine is essential to the viability of the embryo. The embryo is dependent on a maternal source of the amino acid. For embryo lethality to occur, a significant reduction of maternal glutamine is required. Such reduction in maternal glutamine depends on a significant inhibition of glutamine synthetase by GA. This can only occur when the mother is exposed to very high levels of GA. SPECIFICALLY: a. The reproductive toxicity of GA is confined to very short, early stages of reproduction, during which the conceptus is dependent on maternal glutamine; and b. In order for the effect to occur, significant reduction in maternal blood glutamine level is required, which in turn depends on a significant inhibition of glutamine synthetase, induced by high levels of GA in the maternal system. There is no evidence for accumulation of GA in the mammalian organism beyond a factor of two and no evidence for its metabolic toxification. To raise a concern in humans, women would have to be exposed to GA during the very limited time frame of preimplantation or implantation and the exposure would have to be to the exceedingly high levels necessary to alter the maternal metabolism and, correspondingly, result in glutamine levels in maternal tissue and blood plasma being drastically reduced. There is no basis to suggest that such exposures would occur under conditions of normal handling and use. SPECIFICALLY: a. Under conditions of normal handling and use, operators would never be exposed to GA levels that could potentially inhibit glutamine synthetase to the extent that this inhibition could impair preimplantation or implantation. b. All acceptable exposure measurements and predictive calculations confirm this conclusion, and in fact demonstrate that reasonably foreseeable exposure of workers would be to levels significantly below the AOEL. c. The evidence is also clear that there is no reproductive toxicity hazard to workers upon reentry tosprayed fields, bystanders, consumers or toddlers. The safety margin compared to the NOAEL in animal studies is sufficiently large to assure protection of the health of workers using GA as well as bystanders, consumers, and toddlers. Pursuant to Annex 6 of Commission Directive 2001/59/EC (28th ATP of Council Directive 67/548/EEC), to justify a classification of category 2 there must be sufficient evidence to produce a strong presumption that human exposure to the substance may result in impaired fertility in humans. It is the conclusion of the Science Partners Evaluation Group that there is no reasonable evidence to suggest a strong presumption of impairment. To the contrary, there is clear evidence demonstrating a strong presumption that exposure to GA would not cause the adverse effect demonstrated in rats. Pursuant to Annex 6 of Commission Directive 2001/59/EC (28th ATP of Council Directive 67/548/EEC), to justify a classification of category 3, there must be sufficient evidence to provide a strong suspicion of impaired fertility in humans. There is no basis to conclude that the animal data demonstrating impaired preimplantation or implantation has any relevance to humans in that the effect found in rats only occurs at levels which would never be experienced by workers under conditions of normal handling and use or by bystanders, consumers, or toddlers.

Analysis of dose-response effects on gene expression data with comparison of two microarray platforms

MOTIVATION

The problems of analyzing dose effects on gene expression are gaining attention in biomedical research. A specific challenge is to detect genes with expression levels that change according to dose levels in a non-random manner, but nonetheless may be considered as potential biomarkers.

METHOD

We are among the first to formally apply a tool that uses an isotonic (monotonic) regression approach to this area of study. We introduce a test statistic to select genes with significant dose-response expression in a monotonic fashion based on a permutation procedure. We then compare the results with those achieved from the application of a likelihood ratio-based test.

RESULTS

We apply the isotonic regression approach to a study of gene expression in the RKO colon carcinoma cell line in response to varying dosage levels of the chemotherapeutic agent 5-fluorouracil. A feature of both Affymetrix and printed 75mer oligomer cDNA arrays produced from the same samples provides an opportunity to compare the two microarray platforms.

AVAILABILITY

Statistical software S-plus Code to implement the method is available from the authors.

CONTACT

kcoombes@mdanderson.org

Use of mode of action in risk assessment: past, present, and future

The evolution of chemical risk assessment has been marked by a steadily increasing expectation for the use of chemical-specific dosimetric and mechanistic information to tailor the risk assessment approach. The information to be used can range from the broad physical properties of the chemical to detailed information on the mechanism by which it causes a particular toxic outcome, and the risk assessment decisions effected can in turn range from how to define equivalent exposures across species to whether a particular animal outcome is relevant to a human health assessment. A concept that has proven useful in support of these considerations is the "mode of action," a term coined by the USEPA in their new guidelines for carcinogen risk assessment. This paper describes the increasing use of mode-of-action considerations in risk assessment, beginning with early examples involving quantitative dosimetry on the one hand, and qualitative relevance on the other, which foreshadowed the current interest in mode of action. It then describes more recent developments regarding the use of the mode-of-action concept for the selection of a low-dose extrapolation approach, for harmonization of cancer and noncancer risk assessment approaches, and for cross-chemical evaluations. Finally, examples of recent controversies associated with the use of mode-of-action information in risk assessment are provided to demonstrate the challenges that must be overcome to assure the continued viability of the mode-of-action approach.

Assessing biomarker use in risk assessment--a survey of practitioners

Advances in molecular epidemiology and mechanistic toxicology have provided increased opportunities for incorporating biomarkers in the human health risk assessment process. For years, the published literature has lauded the concept of incorporating biomarkers into risk assessments as a means to reduce uncertainty in estimating health risk. For all the potential benefits, one would think that markers of effective dose, markers of early biological effects, and markers of human susceptibility are frequently selected as the basis for quantitative human health risk assessments. For this article, we sought to determine the degree to which this evolution in risk assessment has come to pass. The extent to which biomarkers are being used in current human health risk assessment was determined through an informal survey of leading risk assessment practitioners. Case studies highlighting the evolution of risk assessment methods to include biomarkers are also described. The goal of this review was to enhance the implementation of biomarker technology in risk assessment by (1) highlighting successes in biomarker implementation, (2) identifying key barriers to overcome, and (3) describing evolutions in risk assessment methods.

Whole body pharmacokinetic models

The aim of the current review is to summarise the present status of physiologically based pharmacokinetic (PBPK) modelling and its applications in drug research, and thus serve as a reference point to people interested in the methodology. The review is structured into three major sections. The first discusses the existing methodologies and techniques of PBPK model development. The second describes some of the most interesting PBPK model implementations published. The final section is devoted to a discussion of the current limitations and the possible future developments of the PBPK modelling approach. The current review is focused on papers dealing with the pharmacokinetics and/or toxicokinetics of medicinal compounds; references discussing PBPK models of environmental compounds are mentioned only if they represent considerable methodological developments or reveal interesting interpretations and/or applications.The major conclusion of the review is that, despite its significant potential, PBPK modelling has not seen the development and implementation it deserves, especially in the drug discovery, research and development processes. The main reason for this is that the successful development and implementation of a PBPK model is seen to require the investment of significant experience, effort, time and resources. Yet, a substantial body of PBPK-related research has been accumulated that can facilitate the PBPK modelling and implementation process. What is probably lagging behind is the expertise component, where the demand for appropriately qualified staff far outreaches availability.

A physiologically based pharmacokinetic model for oxytetracycline residues in sheep

A physiologically based pharmacokinetic model (PBPK) for oxytetracycline (OTC) residues in sheep was developed using previously published data from a combined serum pharmacokinetic and tissue residue study [Craigmill et al. (2000) J. Vet. Pharmacol. Ther.23, 345]. Physiological parameters for organ weights and tissue blood flows were obtained from the literature. The tissue/serum partition coefficients for OTC were estimated from the serum and tissue residue data obtained at slaughter. The model was developed to include all of the tissues for which residue data were available (serum, kidney, liver, fat, muscle and injection site), and all of the remaining tissues were combined into a slowly perfused compartment with low permeability. Total body clearance of OTC calculated in the previous study was used as the starting value for clearance in the PBPK model, with the kidney being the only eliminating organ. The model was built using ACSL (Advanced Continuous Simulation Language) Graphic Modeler, and the model was fit to the serum and tissue data using the ACSL Math/Optimizer software (AEgis Technologies Group, Inc., Huntsville, AL, USA). A sensitivity analysis was also performed to determine which parameters had the greatest effect on the goodness of fit. Numerous strategies were tested to model the injection site, and a model providing a biexponential absorption of the drug from the injection bolus gave the best fit to the experimental data. The model was validated using the clearance parameters calculated from the traditional pharmacokinetic model for each individual animal in the PBPK model. This simple PBPK model well predicted OTC residues in sheep tissues after intramuscular dosing with a long-acting preparation and may find use for other species and other veterinary drugs.

Impact of physiologically based pharmacokinetic modeling on benchmark dose calculations for TCDD-induced biochemical responses

In risk assessment, noncancer risk is currently estimated using a no observed adverse effect level (NOAEL) from an experimental dose-response study, divided by uncertainty factors, to estimate a presumably safe level of human exposure. A benchmark dose approach, in which an effective dose (ED) resulting in a specified percentage increase over background for effects is estimated by empirical modeling, has been proposed as a replacement for the NOAEL methodology. The aim of this analysis is to compare methods for estimation of body burden resulting in a 1 or 10% maximum increase over background (BB(01) or BB(10)) for biochemical responses following exposure to 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) in female Sprague-Dawley rats. In one method, an ED resulting in a prespecified increase in response over background was estimated using average daily doses and an empirical Hill model. The ED was then converted to an equivalent body burden by a simple kinetic model assuming steady-state conditions, half-life of TCDD in the rat, and 100% absorption of TCDD. Alternatively, a mechanistic physiologically based pharmacokinetic (PBPK) model of TCDD in the rat was used to predict body burdens for administered doses. These PBPK-modeled body burdens were then used directly by the Hill model to calculate a BB(01) or BB(10). In general, the body burden values derived from EDs were within five-fold of BB(01) or BB(10) calculated from the PBPK model. BB(01) and BB(10) values from both methods were within two orders of magnitude of current human general population exposure to all dioxin-like compounds.

DNA microarrays and toxicogenomics: applications for ecotoxicology?

Toxicogenomics attempts to define how the regulation and expression of genes mediate the toxicological effects associated with exposure to a chemical. DNA microarrays are rapidly becoming one of the tools of choice for large-scale toxicogenomic studies. An approach in modern toxicogenomics has been to classify toxicity based on gene transcriptional patterns; comparing the transcriptional responses of a chemical with unknown toxicity to those for which the transcriptional profiles and toxicological endpoints have been well characterized. Recent evidence suggests that gene expression microarrays may be instrumental in defining mechanisms of action of toxicants. However, several assumptions are inherent to a toxicogenomic-based approach in toxicology, many of which remain to be validated. Gene expression profiling using DNA microarrays represents a snapshot of the gene transcriptional responses occurring at a particular time and within a particular tissue. Toxicity, on the other hand, represents a continuum of possible effects governed by both temporal and spatial factors that are inextricably contingent upon the exposure conditions. The perceived toxicological properties of any chemical are dependent on the route, dose, and duration of the exposure, and as such, gene expression patterns are also subject to these variables. Correct interpretation of DNA microarray data for the assessment of the toxicological properties of chemicals will require that temporal and spatial gene expression profiles be accounted for. These considerations are further compounded in ecotoxicological studies, during which altered gene expression patterns induced from exposure to an anthropogenic substance must be discernible over and above the complex effects that phenotypic, genotypic, and environmental variables have on gene expression. To this end, the greatest utility of DNA microarrays in the field of ecotoxicology may be in predicting the toxicological modes of action of anthropogenic substances on host physiology, particularly in non-model organisms. Predictable and accurate assessment of the impacts of a chemical substance in ecotoxicology will require that classical toxicological endpoints be used to validate any effects predicted based on gene expression profiling. Validated expression profiling may subsequently find utility in ecotoxicological-based computer simulation models, such as the Biotic Ligand Model (BLM), in which gene expression information may be integrated with geochemical, pharmacokinetic, and physiological data to accurately assess and predict toxicity of metals to aquatic organisms.