The basic assumption for QSAR study is that molecules that have similar structures tend to have similar properties. The objective of a QSAR model is to relate a group of predictor variables with the potency of the responsive variable. The predictors contain the information that describes theoretical molecular descriptors or physicochemical properties of chemicals; the responsive variables include various properties, such as biological activity, that are used for the classification of hazardous chemicals. The development of descriptors from molecular structure and correlating molecular descriptors with responsive activities through multivariate analysis are two steps to establish QSAR models. The process of developing QSAR models is not covered in this chapter.

The predictors contain chemical information encoded in the molecular structure from a practical perspective, and the chemical information includes constitutional, geometrical, hydrophobic, electronic, steric, and topological descriptors. More information regarding the descriptor can be found in the literature (Helguera et al., 2008a; Todeschini and Consonni, 2000). The descriptors can be obtained using the quantum chemical software. Once it is determined by the software, it can be considered as an independent variable to develop QSAR models. List of quantum chemical softwares is summarized in literature (Nantasenamat et al., 2009).

Before the implementation of multivariate analysis, data preprocessing is an essential step to ensure the integrity of the dataset contributing to QSAR studies, which usually contains data cleaning, data transformation, and feature selection. The multivariate analysis correlates predictors with responsive variables through mathematical regression. Linear regression is a common approach to establishing the correlation between the responsive variables and a number of the chemical predictors. For cases in which the correlations cannot be explained by a linear relation, nonlinear methods are used for multivariate analysis to develop QSAR models. Some of the popular nonlinear techniques include artificial neural network (Nantasenamat et al., 2009), partial least squares regression (Wold et al., 2001), and support vector machine (Wang, 2005). The developed QSAR models require validation and evaluation of the performance in predicting desirable properties through statistical methods. The validation is conducted using statistical parameters, such as Pearson's correlation coefficient (r) and root-mean-square (RMS) error, as well as F-test and outliers. A commonly adopted method to evaluate the ability of developed QSAR models to predict properties of interest was given in Tropsha et al.'s recommendation (Tropsha et al., 2003).

QSAR has been used to predict different health hazards. Based on the criteria specified in Appendix A of the Hazard Communication Standard (29 CFR 1910. 1200) promulgated by the Occupational Safety and Health Administration (OSHA), the health hazards of chemicals are categorized by GHS classification as acute toxicity, skin corrosion and irritation, serious eye damage and eye irritation, respiratory or skin sensitization, germ cell mutagenicity, carcinogenicity, reproductive toxicity, specific target organ toxicity due to single exposure or repeated/prolonged exposure, and aspiration hazard (OSHA, 2014).

Acute toxicity refers to “those adverse effects occurring following oral or dermal administration of a single dose of a substance, or multiple doses given within 24 h, or an inhalation exposure of 4 h” (OSHA, 2014).

Acute toxicity is characterized by LD₅₀ for oral and dermal intake, LC₅₀ for inhalation, or the acute toxicity estimates. QSAR studies have been conducted to predict LD₅₀, LC₅₀, or EC₅₀ for many chemicals, including pesticides, benzene derivatives, aliphatic compounds, alcohol ethoxylates, and chemical mixtures (Devillers and Flatin, 2000; Gute and Basak, 1997; Croni et al., 2000; Morrall et al., 2003; Wei et al., 2004). QSAR studies were also conducted to predict acute aquatic toxicity (Gute and Basak, 1997), which is not covered by OSHA, since it does not have the authority to regulate environmental issues.

The QSAR models predicting mutagenicity and carcinogenicity have been reviewed by Romualdo Benigni (Tatiana et al., 2007; Benigni and Bossa, 2008).

A mutation is defined as “a permanent change in the amount or structure of the genetic material in a cell. The term mutation applies both to heritable genetic changes that may be manifested at the phenotypic level and to the underlying DNA modifications when known (including, for example, specific base pair changes and chromosomal translocations). The term mutagenic and mutagen will be used for agents giving rise to an increased occurrence of mutations in populations of cells and/or organisms”(OSHA, 2014).

Carcinogen means “a substance or a mixture of substances which induce cancer or increase its incidence. Substances and mixtures which have induced benign and malignant tumors in well-performed experimental studies on animals are considered also to be presumed or suspected human carcinogens unless there is strong evidence that the mechanism of tumor formation is not relevant for humans” (OSHA, 2014).

Tatiana et al. (2007) evaluated 78 noncommercial QSAR models through two-phase analysis. Benigni and Bossa (2008) reviewed 13 local QSAR models, which concluded that 30–70% were correct for potency-generated predictions and 70–100% were correct for discriminating between active and inactive chemicals in the external prediction. Some other QSAR models predicting carcinogenicity are summarized in Table 9.1.

QSAR studies were also performed to predict skin corrosion and irritation. Skin corrosion is “the production of irreversible damage to the skin; namely, visible necrosis through the epidermis and into the dermis, following the application of a test substance for up to 4 h. Corrosive reactions are typified by ulcers, bleeding, bloody scabs, and, by the end of observation at 14 days, by discoloration due to blanching of the skin, complete areas of alopecia, and scars” (OSHA, 2014). Skin irritation is “the production of reversible damage to the skin following the application of a test substance for up to 4 h” (OSHA, 2014). The corrosivity has been correlated with log P_ow (logarithm of octanol/water partition coefficient), melting point, molecular volume, and pK_a, since they characterize either skin permeability or cytotoxicity. A lower log P_ow value, either from large molecular volumes or high melting points, indicates the tendency of low skin permeability for chemicals, that is, less corrosive, unless the pK_a value shows they are either acidic or basic (Barratt, 1996). Some QSAR studies predict irritation by calculating the primary irritation index (PII) or skin irritation score, which are correlated with log P_ow, molecular volume, the dipole moment, or the molecular weight.

QSAR studies use structural fragments and chemical substructures to predict eye irritation/corrosion, and the QSAR models are reviewed by Saliner et al. (2008). Serious eye damage is “the production of tissue damage in the eye, or serious physical decay of vision, following application of a test substance to the anterior surface of the eye, which is not fully reversible within 21 days of application” (OSHA, 2014). Eye irritation is “the production of changes in the eye following the application of test substance to the anterior surface of the eye, which are fully reversible within 21 days of application” (OSHA, 2014).

Total erythema score (TES) and EC3 are simulated for the prediction of respiratory and skin sensitizer. Respiratory sensitizer means “a chemical that will lead to hypersensitivity of the airways following inhalation of the chemical. Skin sensitizer means a chemical that will lead to an allergic response following skin contact” (OSHA, 2014). TES is used to indicate the index of biological property and EC3 is considered as a quantitative measure of sensitizing potential (Barratt et al., 1994; Patlewicz et al., 2003).

QSAR model has also been developed to predict reproductive toxicity (Hewitt et al., 2007). In this QSAR model, 184 heterogeneous antivirals and tocolytics compounds were evaluated for placental clearance index or placental transfer index. Reproductive toxicity includes “adverse effects on sexual function and fertility in adult males and females, as well as adverse effects on development of the offspring. Some reproductive toxic effects cannot be clearly assigned to either impairment of sexual function and fertility or to developmental toxicity. Nonetheless, chemicals with these effects shall be classified as reproductive toxicants” (OSHA, 2014).

Currently, there are no QSAR studies to predict toxicity to specific target organs for single or repeated exposures in the literature (Quintero et al., 2012). “Specific target organ toxicity-single exposure (STOT-SE), means specific, non-lethal target organ toxicity arising from a single exposure to a chemical.” “Specific target organ toxicity-repeated exposure (STOT-RE), means specific target organ toxicity arising from repeated exposure to a substance or mixture” (OSHA, 2014).

No aspiration hazard prediction using the QSAR method has been found in the literature. Aspiration toxicity includes “severe acute effects such as chemical pneumonia, varying degrees of pulmonary injury or death following aspiration” (OSHA, 2014).