Toxicology QSARs

If in-depth toxicological QSAR is to be performed on a series of candidates, it is likely that new data and modeling will be required. 

Doucette, W.J., and M.S. Holt. 1992. PEP, A microcomputer program for estimating physical/chemical properties. In Fifth International Workshop on QSAR in Environmental Toxicology (QSAR 92). Duluth, MN, July 19-23. 

Schultz TW, Hewitt M, Netzeva TI, Cronin MTD (2007) Assessing applicability domains of toxicological QSARs definition, confidence in predicted values, and the role of mechanisms of action. QSAR Comb Sci 26 238-254... 

Central to the issues of quality, transparency, and domain identification as they relate to toxicological QSAR is biological data. High quality toxicity data on a structurally diverse set of molecules are required to formulate and validate high quality QSARs. Quality toxicity data typically come from standardized assays measured in a consistent manner, with a clear and unambiguous endpoint, and low experimental error. In such cases, quality is associated with values, which are accurate, consistent with other data within the same set, and consistent with data for other similar endpoints. In the case of comparisons between endpoints, it is as important for data to be consistent between endpoints as for the inconsistencies to be consistent. 

The primary supposition of any toxicological QSAR is that the potency of a compound is dependent upon its molecular structure, which is typically quantified by chemical properties (Schultz et al., 2002). Chemical descriptors include a variety of types, including atom, substituent, and molecular parameters. The most transparent of these are the molecular-based empirical and quantum chemical descriptors. Empirical descriptors are measured descriptors and include physicochemical properties such as hydrophobicity (Dearden, 1990). Quantum chemical properties are theoretical descriptors and include charge and energy values (Karelson et al., 1996). Physicochemical and quantum chemical descriptors are for the most part easily interpretable with regard to how that property may be related to toxicity. The classic example of this, the partitioning of a toxicant between aqueous and lipid phases, has been used as a measure of hydrophobicity for over a century (Livingstone, 2000). 

As of April 2006, the preliminary REACH technical guidance documents (TGD) available to date relate to the risk assessment processes of REACH, such as the use of toxicological QSARs and the development of exposure scenarios [126]. 

This chapter describes the algorithms of the various data analysis methods currently used for developing toxicological QSAR models. Data collection, data pre-processing, computation and selection of molecular descriptors, and model validation have been extensively reviewed elsewhere [2-11], so they are not described here. Freely available online software and commercial software available for constructing QSAR models of various toxicological properties prediction are also discussed. 

A disadvantage of the majority of these software is that they do not have the capability to calculate molecular descriptors. Thus additional software, such as DRAGON [71], Molconn-Z [72], or MODEL [73], are needed to enable toxicological QSAR models to be built. 

Netzeva,T.I., Aptula, A.O., Benfenati, E., Cronin, M.T D., Gini, G., Lessigiarska, L, Maran, U., Vracko, M. and Schiitirmann, G. (2005) Description of the electronic structure of organic chemicals using semiempirical and ah initio methods for development of toxicological QSARs./. Chem. Inf. Model, 45, 106-114. 

Ovahty descriptors presented here have been applied to study the toxic activity of various compounds including alcohols, esters, carboxylic acids, amines (Vlaia et al. 2009 Vlaia 2010). Here we present a toxicological QSAR study realized on a series of 28 amines. 

In (eco)toxicological QSAR studies, the molecular descriptor of choice is the n-octanol/water partition coefficient (log P), generally used in a simple regression equation. However, sometimes a simple linear regression model is inadequate to model properly the dependence of biological activity (BA) on logF. For example, fish exposed to very hydrophobic chemicals for a limited test duration have insufficient time to achieve a pseudo-steady state partitioning equilibrium between the toxicant concentration in aqueous circumambient phase and the hydrophobic site of action within the fish. Hansch initiated the use of a parabolic model in log P (equation 1) to overcome... 

---