Best Practices for QSAR Model Reporting: Physical and Chemical Properties, Ecotoxicity, Environmental Fate, Human Health, and Toxicokinetics Endpoints.

BACKGROUND: Quantitative and qualitative structure­activity relationships (QSARs) have been used to understand chemical behavior for almost a century. The main source of QSAR models is the scientific literature, but the open question is how well these models are documented. OBJECTIVES: The main aim of this study was to critically analyze the publication practices of QSARs with regard to transparency, potential reproducibility, and independent verification. The focus was on the level of technical completeness of the published QSARs. METHODS: A total of 1,533 QSAR articles reporting 79 individual endpoints, mostly in environmental and health science, were reviewed. The QSAR parameters required for technical completeness were grouped into five categories: chemical structures, experimental endpoint values, descriptor values, mathematical representation of the model, and predicted endpoint values. The data were summarized and discussed using Circos plots. RESULTS: Altogether, 42.5% of the reviewed articles were found to be potentially reproducible. The potential reproducibility for different endpoint groups varied; the respective rates were 39% for physical and chemical properties, 52% for ecotoxicity, 56% for environmental fate, 30% for human health, and 32% for toxicokinetics. The reproducibility of QSARs is discussed and placed in the context of the reproducibility of the experimental methods. Included are 65 references to open QSAR datasets as examples of models restored from scientific articles. DISCUSSION: Strikingly poor documentation of QSARs was observed, which reduces the transparency, availability, and consequently, the application of research results in scientific, industrial, and regulatory areas. A list of the components needed to ensure the best practices for QSAR reporting is provided, allowing long-term use and preservation of the models. This list also allows an assessment of the reproducibility of models by interested parties such as journal editors, reviewers, regulators, evaluators, and potential users. https://doi.org/10.1289/EHP3264.

Topological Fingerprints as an Aid in Finding Structural Patterns for LRRK2 Inhibition.

Multiplet-based fingerprint mapping has been used to analyse the relationship between the structural features of potential drug candidates and the enzyme LRRK2 inhibition expressed as the inhibition constant (pKi ). For 198 structurally diverse compounds 4195 dimensional fingerprints were generated and mathematically manipulated using partial least squares (PLS) regression. A variation of PLS-BETA technique was developed for the reduction of noise by eliminating excess variables that resulted in a 636 dimensional fingerprint related to pKi . The QSAR model for the training set of 170 compounds (R(2) =0.87, Q(2) =0.77 and SDEC=0.42) had four latent variables (PLS components) and it was validated by the external test set of 28 compounds (Qext (2) =0.63). The proposed model of LRRK2 inhibitory activity can be helpful in designing focused libraries enriched in LRRK2 inhibitors and identifying new active chemotypes in compound databases.

Quantitative structure-activity relationship (QSAR) modeling of EC50 of aquatic toxicities for Daphnia magna.

The experimental EC(50) toxicities toward Daphnia magna for a series of 130 benzoic acids, benzaldehydes, phenylsulfonyl acetates, cycloalkane-carboxylates, benzanilides, and other esters were studied using the Best multilinear regression algorithm (BMLR) implemented in CODESSA. A modified quantitative structure-activity relationships (QSAR) procedure was applied guaranteeing the stability and reproducibility of the results. Separating the initial data set into training and test subsets generated three independent models with an average R(2) of .735. A five-descriptor general model including all 130 compounds, constructed using the descriptors found effective for the independent subsets, was characterized by the following statistical parameters: R(2) = .712; R(2)(cv) = .676; F = 61.331; s(2) = 0.6. The removal of two extreme outliers improved significantly the statistical parameters: R(2) = .759; R(2)(cv) = .728; F = 77.032; s(2) = 0.499. The sensitivity of the general model to chance correlations was estimated by applying a scrambling procedure involving 20 randomizations of the original property values. The resulting R(2) = .192 demonstrated the high robustness of the model proposed. The descriptors appearing in the obtained models are related to the biochemical nature of the adverse effects. An additional study of the EC(50)/LC(50) relationship for a series of 28 compounds (part of our general data set) revealed that these endpoints correlated with R(2) = .98.

Improved efficiency of focal point conformational analysis with truncated correlation consistent basis sets.

It has been suggested that the computational cost of correlated ab initio calculations could be reduced efficiently by using truncated basis sets on hydrogen atoms (Mintz et al., J Chem Phys 2004, 121, 5629). We now explore this proposal in the context of conformational analysis of small molecules, such as hydrogen peroxide, dimethyl ether, ethyl methyl ether, formic acid, methyl formate, and several small alcohols. It is found that truncated correlation consistent basis sets that lack certain higher angular momentum functions on hydrogen atoms offer accuracy similar to traditional Dunning's basis sets for conformational analysis. Combination of such basis sets with the basis set extrapolation technique to estimate Hartree-Fock and Møller-Plesset second order energies provides composite extrapolation model chemistries that are significantly more accurate and faster than analogous single point calculations with traditional correlation consistent basis sets. Root mean square errors of best composite extrapolation model chemistries on the used set of molecules are within 0.03 kcal/mol of traditional focal point conformational energies. The applicability of composite extrapolation methods is illustrated by performing conformational analysis of tert-butanol and cyclohexanol. For comparison, conformational energies calculated with popular molecular mechanics force fields are also given.

Modeling the toxicity of chemicals to Tetrahymena pyriformis using heuristic multilinear regression and heuristic back-propagation neural networks.

During the last years, considerable effort has been devoted to model the toxicity of chemicals to Tetrahymena pyriformis for medium and large sized data sets using various artificial neural network (ANN) techniques. Motivation behind this has been to model highly complex relationships with nonlinear character making it possible to describe wide structural diversity within one model. The current work compares the performance of two heuristic methods in developing quantitative structure-activity relationship (QSAR) models: the best multilinear regression (BMLR) approach and the heuristic back-propagation neural networks (hBNN). The modeling is based on a diverse data set of 1371 organic chemicals with toxicity data (log(1/IGC50)) collected from the literature. The toxicity values correspond to the static 40-h Tetrahymena pyriformis population growth impairment assay. The comparison of the two methods showed that the BMLR approach produces acceptable QSAR models (R2 = 0.726), whereas the hBNN method produced a statistically more significant model (R2 = 0.826) for the given endpoint. The hBNN method was able to relate different descriptors to the toxicity than the BMLR method. Both models were validated with an external prediction set. The descriptors in the models were analyzed and discussed.

Comparative quantitative structure-activity-activity relationships for toxicity to Tetrahymena pyriformis and Pimephales promelas.

An approach for predicting acute aquatic toxicity, in the form of a quantitative structure-activity-activity relationship (QSAAR), is described. This study assessed relative toxic effects to a fish, Pimephales promelas, and a ciliate, Tetrahymena pyriformis, and attempted to form relationships between them. A good agreement between toxic potencies (R2 = 0.754) was found for a chemically diverse dataset of 364 compounds, when using toxicity to the ciliate as a surrogate to that for fish. This relationship was extended by adding three theoretical structural descriptors of the molecules. The inclusion of these descriptors improved the relationship further (R2 = 0.824). The structural features that were found to improve the extrapolation between the toxicity to the two different species were related to the electron distribution of the carbon skeleton of the toxicant, its hydrogen-bonding ability, and its relative nitrogen content. Such a QSAAR approach provides a potential tool for predicting the toxicities of chemicals for environmental risk assessment and thus for reducing animal tests.

QSPR treatment of the soil sorption coefficients of organic pollutants.

In this study, general and class-specific QSPR models for soil sorption, logK(OC), of 344 organic pollutants (0 < logK(OC) < 4.94) were developed using a large variety of theoretical molecular descriptors based only on molecular structure. Two general models were obtained. The first model was derived for a structurally representative set of 68 chemicals (R2=0.76, s=0.44), whereas the second involved a total of 344 compounds (R2=0.76, s=0.41). The first was validated using the data for the remaining 276 pollutants (R2=0.70, s=0.45). An additional validation of both models was performed using an independent set of 48 pollutants. Both models predict the logK(OC) at the level of experimental precision, while the theoretical molecular descriptors appearing in the QSPR models give further insight into the mechanisms of soil sorption. The analysis of the distribution of the residuals of the logK(OC) values calculated by both general models indicated the need and possible advantages of modeling soil sorption for smaller data sets related to individual classes of chemicals. Accordingly, QSPR models were also developed for 14 chemical classes. The descriptors appearing in these models were discussed as related to the possible interaction mechanisms in soil sorption.