An approach for evaluating and increasing the informational content of mutagenicity and clastogenicity data bases

Similarities in the Mechanisms of Antibacterial Activity (Microtox™ Assay) and Toxicity to Vertebrates

The structural determinants associated with inhibition of bacterial growth (Microtox assay), as well as those associated with other toxicological activities, were identified using the CASE/Multi-CASE structure-activity relational system. CASE/Multi-CASE also possesses the ability to compare structural determinants associated with different biological phenomena. The extent of the homology between structural determinants can be used to evaluate common mechanisms of action. Using this approach, significant commonalities between structural determinants associated with the Microtox assay and those associated with toxicity to mammals and fish were found, suggesting that there may be a mechanistic relationship between these toxicological activities. The findings indicate that, within these structural similarities, the Microtox assay can be used as an indicator of potential toxicity to mammals and fish.

Database composition can affect the structure-activity relationship prediction

The percent active (A) and inactive (I) chemicals in a database can directly affect the sensitivity (% active chemicals predicted correctly) and specificity (% inactive chemicals predicted correctly) of structure-activity relationship (SAR) analyses. Subdividing the National Center for Toxicological Research (NCTR) liver cancer database (NCTRlcdb) into various A/I ratios, which varied from 0.2 to 5.5, resulted in sensitivity/specificity ratios that varied from 0.1 to 6.5. As percent active chemicals increased (increasing A/I ratio), the sensitivity rose, the specificity decreased, and the concordance (% total chemicals predicted correctly) remained fairly constant. The numbers of chemicals in the various data sets ranged from 187 to 999 and appeared to have no affect on any of the 3 predictors of sensitivity, specificity, or concordance.

Computer-aided rodent carcinogenicity prediction

The potential of the computer program PASS (Prediction Activity Spectra for Substances) to predict rodent carcinogenicity for chemical compounds was studied. PASS predicts carcinogenicity of chemical compounds on the basis of their structural formula and of structure-activity relationship analysis of known carcinogens and non-carcinogens. The data on structures and experimental results of 2-year carcinogenicity assays for 412 chemicals from the NTP (National Toxicological Program) and 1190 chemicals from the CPDB (Carcinogenic Potency Database) were used in our study. The predictions take into consideration information about species and sex of animals. For evaluation of the predictive accuracy we used two procedures: leave-one-out cross-validation (LOO CV) and leave-20%-out cross-validation. In the last case we randomly divided the studied data set 20 times into two subsets. The data from the first subset, containing 80% of the compounds, were added to the PASS training set (which includes about 46,000 compounds with about 1500 biological activity types collected during the last 20 years to predict biological activity spectra), the second subset with 20% of the compounds was used as an evaluation set. The mean accuracy of prediction calculated by LOO CV is about 73% for NTP compounds in the 'equivocal' category of carcinogenic activity and 80% for NTP compounds in the 'evidence' category of carcinogenicity. The mean accuracy of prediction for the CPDB database is 89.9% calculated by LOO CV and 63.4% calculated by leave-20%-out cross-validation. Influence of incorporation of species and sex data on the accuracy of carcinogenicity prediction was also investigated. It was shown that the accuracy was increased only for data on male animals.

SAR Modelling of Complex Phenomena: Probing Methodological Limitations

The increased acceptance of the use of structure–activity relationship (SAR) approaches to toxicity modelling has necessitated an evaluation of the limitations of the methodology. In this study, the limit of the capacity of the MULTICASE SAR program to model complex biological and toxicological phenomena was assessed. It was estimated that, provided the data set consists of at least 300 chemicals, divided equally between active and inactive compounds, the program is capable of handling phenomena that are even more “complex” than those modelled up to now (for example, allergic contact dermatitis, Salmonella mutagenicity, biodegradability, inhibition of tubulin polymerisation). However, within the data sets currently used to generate SAR models, there are limits to the complexity that can be handled. This may be the situation with regard to the modelling of systemic toxicity (for example, the LD50).

SAR modeling of genotoxic phenomena: the effect of supplementation with physiological chemicals

Structure-activity relationship (SAR) modeling of toxicological phenomena is optimal when the ratio of toxicants to non-toxicants included in the model is unity. Frequently, however, the experimental data available are enriched with toxicants, this appears to be especially true for genotoxicity data sets. It is demonstrated herein, using a Salmonella mutagenicity data set, that when there is a paucity of non-toxicants, the learning set may be augmented with physiological chemicals on the assumption that they are non-genotoxic.

SAR modeling of unbalanced data sets

The increased acceptance of SAR approaches to hazard identification has led us to investigate methods to improve the predictive performance of SAR models. In the present study we demonstrate that although on theoretical grounds the ratio of active to inactive chemicals in the learning set should be unity, SAR models can "tolerate" an unbalanced range in ratios from 3:1 (i.e., 75% actives) to 1:2 (i.e., 33% actives) and still perform adequately. On the other hand SAR models derived from learning sets with ratios in excess of 4:1 (80% actives), even when corrected for the initial ratio do not perform satisfactorily.

Knowledge discovery and data mining in toxicology

Knowledge discovery and data mining tools are gaining increasing importance for the analysis of toxicological databases. This paper gives a survey of algorithms, capable to derive interpretable models from toxicological data, and presents the most important application areas. The majority of techniques in this area were derived from symbolic machine learning, one commercial product was developed especially for toxicological applications. The main application area is presently the detection of structure-activity relationships, very few authors have used these techniques to solve problems in epidemiological and clinical toxicology. Although the discussed algorithms are very flexible and powerful, further research is required to adopt the algorithms to the specific learning problems in this area, to develop improved representations of chemical and biological data and to enhance the interpretability of the derived models for toxicological experts.

Modeling the mouse lymphoma forward mutational assay: the Gene-Tox program database

An SAR model of the induction of mutations at the tk(+/-) locus of L5178Y mouse lymphoma cells (MLA, for mouse lymphoma assay) was derived based upon a re-evaluation of experimental results reported by a Gene-Tox (GT) working group [A.D. Mitchell, A.E. Auletta, D. Clive, P.E. Kirby, M.M. Moore, B.C. Myhr, The L5178Y/tk(+/-) mouse lymphoma specific gene and chromosomal mutation assay. A phase III report of the U.S. Environmental Protection Agency Gene-Tox Program, Mutation Res. 394 (1997) 177-303.]. The predictive performance of the GT MLA SAR model was similar to that of a Salmonella mutagenicity model containing the same number of chemicals. However, the structural determinants (biophores) derived from the GT MLA SAR model include both electrophilic as well as non-electrophilic moieties, suggesting that the induction of mutations in the MLA may occur by both direct interaction with DNA and by non-DNA-related mechanisms. This was confirmed by the observation that the set of biophores associated with MLA overlapped significantly with those associated with phenomena related to loss of heterozygosity, chromosomal rearrangements and aneuploidy. The MLA SAR model derived from the GT data evaluation was significantly more predictive than an SAR model previously derived from MLA data reported by the US National Toxicology Program [B. Henry, S.G. Grant, G. Klopman, H.S. Rosenkranz, Induction of forward mutations at the thymidine kinase locus of mouse lymphoma cells: evidence for electrophilic and non-electrophilic mechanisms, Mutation Res. 397 (1998) 331-335.]. Moreover, the latter model appeared to be more complex than the former, suggesting that the GT induction data was both simpler mechanistically and more homogeneous than that of the NTP.

Structural determinants of developmental toxicity in hamsters

A CASE/MULTICASE structure activity relationship (SAR) model of developmental toxicity of chemicals in hamsters (HaDT) was developed. The model exhibited a predictive performance of 74%. The model's overall predictivity and informational content were similar to those of an SAR model of mutagenicity in Salmonella. However, unlike the Salmonella mutagenicity model, the HaDT model did not identify overtly chemically reactive moieties as associated with activity. Moreover, examination of the number and nature of significant structural determinants suggested that developmental toxicity in hamsters was not the result of a unique mechanism or attack on a specific molecular target. The analysis also indicated that the availability of experimental data on additional chemicals would improve the performance of the SAR model.

Development, characterization and application of predictive-toxicology models

The adoption of SAR techniques for risk assessment purposes requires that the predictive performance of models be characterized and optimized. The development of such methods with respect to CASE/MULTICASE are described. Moreover, the effects of size, informational content, ratio of actives/inactives in the model on predictivity must be determined. Characterized models can provide mechanistic insights: nature of toxicophore, reactivity, receptor binding. Comparison of toxicophores among SAR models allows a determination of mechanistic overlaps (e.g., mutagenicity, toxicity, inhibition of gap junctional intercellular communication vs. carcinogenicity). Methods have been developed to combine SAR submodels and thereby improve predictive performance. Now that predictive toxicology methods are gaining acceptance, the development of Good Laboratory Practices is a further priority, as is the development of graduate programs in Computational Toxicology to adequately train the needed professional.

Induction of forward mutations at the thymidine kinase locus of mouse lymphoma cells: evidence for electrophilic and non-electrophilic mechanisms

A database of 209 chemicals tested for induction of forward mutations at the heterozygous thymidine kinase (TK +/-) locus in L5178Y mouse lymphoma cells was analyzed for structure-activity relationships using the MultiCASE expert system. Consistent with evidence of several contributing biological mechanisms, these studies suggest that such mutations may occur by more than one mechanism. As might be expected, there was evidence for a component involving direct electrophilic attack on the cellular DNA, in a manner previously established as causative in the induction of mutations in Salmonella. In addition, however, there was also strong evidence for another mechanism or mechanisms involving chromosome missegregation, cellular toxicity or an alternate site of action, such as the microtubules.

Structural and mechanistic bases for the induction of mitotic chromosomal loss and duplication ('malsegregation') in the yeast Saccharomyces cerevisiae: relevance to human carcinogenesis and developmental toxicology

MultiCASE has the ability to automatically determine the structural features responsible for the biological activity of chemicals. In the present study, 93 chemicals tested for their ability to induce chromosomal 'malsegregation' in the yeast Saccharomyces cerevisiae were analyzed. This 'malsegregation' mimics molecular events that occur during human development and carcinogenesis resulting in an effective loss of one chromosome of an autosomal pair and duplication of the homologue. Structural features associated with the ability to induce such chromosome loss and duplication were identified and compared with those obtained from examination of other toxicological data bases. The most significant structural similarities were identified between the induction of chromosomal malsegregation and several toxicological phenomena such as cellular toxicity, induction of sister chromatid exchanges in vitro and rodent developmental toxicity. Very significant structural similarities were also found with systemic toxicity, induction of micronuclei in vivo and human developmental toxicity. Less significant structural overlaps were found between yeast malsegregation and rodent carcinogenicity, DNA reactivity and mutagenicity, and the induction of chromosome aberrations in vitro and sister chromatid exchanges in vivo. These overlaps may indicate mechanistic similarities between the induction of chromosomal malsegregation and other toxicological phenomena. The predictivity of the SAR model derived from the present data base is relatively low, however. This may be merely a reflection of the small size and composition of the data base, however, further analyses suggest that it reflects primarily the multiple mechanisms responsible for the induction of chromosomal malsegregation in yeast and the complexity of the phenomenon.

Estimation of the optimal data base size for structure-activity analyses: the Salmonella mutagenicity data base

In the present study, the effects of data base size on predictivity, informational content and structural overlap of derived Structure-Activity Relationship (SAR) models were investigated. It was found that indices of predictivity (i.e., sensitivity, specificity, and concordance between experimental and predicted results (OCP) increased with increasing size of the data base until the range is 300-400 chemicals, at which point they plateau. The greater the size of the data base, the greater the informational content of the model; however, the rate of this increase is no longer optimal when the size of the data base exceeds 400 chemicals.

Structural basis for the induction of preneoplastic glutathione S-transferase positive foci by hepatocarcinogens

A data base consisting of 100 chemicals tested for the ability to enhance the formation of glutathione-S-transferase (GST) positive preneoplastic lesions were analyzed by the CASE structure-activity relational system. A number of structural determinants associated with the induction of GST-positive foci were recognized. The majority of these describe non-electrophilic moieties. It is concluded that there is a structural basis for the induction of these neoplastic lesions; interestingly, it was found that this activity is associated with structures that are non-electrophilic. Reconstruction experiments have indicated that the identified structures are meaningful and that their significance could be better understood with the availability of test results on additional chemicals.