QSAR Assisted Toxicity Prediction

QSAR, as a calculated method has been widely applied in toxicology. QSAR is an acronym for Quantitative Structure-Activity Relationship. The concept was delivered in 1872, and the modern age of QSAR analysis originated from the works of Hansch et al. (1962). By mathematical function, QSAR is a statistical approach which express a relationship between the magnitude of biological effect (BA) and changes in a molecular structure, as Equation (12) showed:

where i denotes a specific chemical of a series, this series may be of homogeneous, or of heterogeneous substances.

In the study of constructing a QSAR model, it is decisive to acquire and screen molecular structure descriptors (also called indicators). At present, more than 3,000 molecular descriptors have been defined and applied to QSAR models to predict group biotoxicity. Choose distinct indicators when adopting the QSAR model that will affect the validity of the predicted results, therefore the preparation of the indicators is the key to the construction (Khan et al., 2020). Among abundant kinds of indicators, the partition coefficient of a chemical between n-octanol and water is supposed to be the most effective indicator (Tichy et al., 2008). However, it is difficult to use K_ow to determine the combined toxicity of mixtures, since the available date was obtained by UV spectrophotometry or HPLC is only suitable for determining the K_ow of single chemicals. Furthermore, Verhaar extended the C₁₈-Empore^TM disks/water partition coefficient (K_MD) to predict the bioconcentration of mixtures, which was found to have a close relationship with logK_ow (Verhaar et al., 1995). The C₁₈-Empore^TM disks/water partition coefficient (K_MD) can be calculated from Equation (13), the unmeasurable problem was successfully solved.

where W is the volume of solution, V is the volume of hydrophobic phase, is the initial amount of chemical i in water, n is the total number of individual chemicals in the mixture, and K_SDi is the partition coefficient of individual chemical i. The value of W/V was suggested to be 6.8 × 10⁵.

Not limited to the partition coefficient of chemical between n-octanol and water, studies concerned other kind indicators to employ QSAR to assess ecotoxicity evaluations of diverse pollutants toward familiar luminescent bacteria, A. fisheri, P. phosphoreum and V. qinghaiensis sp.-Q67, which are categorically reviewed as shown in Table 4.

Characterizations of various types of QSAR models for prediction of joint toxicity.

Researches about QSAR are still of significant interest in the development of innovative models in environmental toxicity prediction. The emergence of the QSAR models fills the gap in predicting the combined toxicity of organic compounds and heavy metals since the information in this field is still scarce. Jin et al. developed three QSARs to determine the individual EC₅₀ of Cd and nine chlorinated anilines (o-chloroaniline, m-chloroaniline, p-chloroaniline, 2,3-dichloroaniline, 2,4-dichloroaniline, 2,5-dichloroaniline, 2,6-dichloroaniline, 3,4-dichloroaniline, and 2,4,5-trichloroaniline) with P. phosphoreum, setting three different levels of Cd concentrations (low, medium and high levels) to mix with chlorinated anilines and the results showed that the number of chlorinated anilines manifesting synergy with Cd is decreasing as the concentration of Cd increases. The robustness of the models was confirmed by comparing the experimental and predicted values, and all the relative error values remain within 16% (Jin et al., 2014). Likewise, the joint toxicities of Cu (low, medium and high levels) with 11 nitroaromatic compounds (nitrobenzene, o-dinitrobenzene, m-nitrobromobenzene, p-nitrobromobenzene, o-nitroaniline, p-nitroaniline, p-nitrobenzoic acid, o-nitrophenol, m-nitrophenol, p-nitrophenol and 2,4-dinitrophenol) were studied by P. phosphoreum with developed QSAR analysis, there is a good agreement between the predicted values and experimental with R² = 0.764, P = 0.000, (Su et al., 2012). By drawing on the study of Su's, Zhang's, Su, Zhang, Li, Qin and Zhang (2019a) work on the acute toxic effect of mixtures between metal Zn and above-mentioned 11 nitroaromatic compounds established robust QSAR models to predict the joint interaction when combined with Zn at low, medium, and high concentrations (Zhang et al., 2019a). Not limited to heavy metals and organics, the joint toxicity of mixed organics can also be predicted by QSAR models. Qin et al. developed a generalized QSAR model for predicting the additive and non-additive toxicities of multi-component mixtures, the experiment tested the joint toxicity of 45 multi-component mixtures composed of two antibiotics (Tetracycline hydrochloric and Chloramphenicol) and four pesticides (Metribuzine, Trichlorfon, dichlorvos, Linuron) to A. fisheri. Compared with classical CA and IA models, the result demonstrated that the QSAR model exhibited high predictive capability for predicting joint toxicity (Qin et al., 2018).