The quality of data used for QSAR model derivation is extremely important as it strongly affects the final robustness and predictive power of the model. Ambiguous or wrong structures need to be carefully checked, because they lead to errors in calculation of descriptors, hence leading to meaningless results. The increasing amounts of data, however, have often made it hard to check of very large databases manually. In the light of this, we designed and implemented a semi-automated workflow integrating structural data retrieval from several web-based databases, automated comparison of these data, chemical structure cleaning, selection and standardization of data into a consistent, ready-to-use format that can be employed for modeling. The workflow integrates best practices for data curation that have been suggested in the recent literature. The workflow has been implemented with the freely available KNIME software and is freely available to the cheminformatics community for improvement and application to a broad range of chemical datasets.

Many neurotoxicants affect energy metabolism in man, but currently available test methods may still fail to predict mito- and neurotoxicity. We addressed this issue using LUHMES cells, i.e., human neuronal precursors that easily differentiate into mature neurons. Within the NeuriTox assay, they have been used to screen for neurotoxicants. Our new approach is based on culturing the cells in either glucose or galactose (Glc–Gal–NeuriTox) as the main carbohydrate source during toxicity testing. Using this Glc–Gal–NeuriTox assay, 52 mitochondrial and non-mitochondrial toxicants were tested. The panel of chemicals comprised 11 inhibitors of mitochondrial respiratory chain complex I (cI), 4 inhibitors of cII, 8 of cIII, and 2 of cIV; 8 toxicants were included as they are assumed to be mitochondrial uncouplers. In galactose, cells became more dependent on mitochondrial function, which made them 2–3 orders of magnitude more sensitive to various mitotoxicants. Moreover, galactose enhanced the specific neurotoxicity (destruction of neurites) compared to a general cytotoxicity (plasma membrane lysis) of the toxicants. The Glc–Gal–NeuriTox assay worked particularly well for inhibitors of cI and cIII, while the toxicity of uncouplers and non-mitochondrial toxicants did not differ significantly upon glucose ↔ galactose exchange. As a secondary assay, we developed a method to quantify the inhibition of all mitochondrial respiratory chain functions/complexes in LUHMES cells. The combination of the Glc–Gal–NeuriTox neurotoxicity screening assay with the mechanistic follow up of target site identification allowed both, a more sensitive detection of neurotoxicants and a sharper definition of the mode of action of mitochondrial toxicants.

Carcinogenicity is a crucial endpoint for the safety assessment of chemicals and products. During the last few decades, the development of quantitative structure–activity relationship ((Q)SAR) models has gained importance for regulatory use, in combination with in vitro testing or expert-based reasoning. Several classification models can now predict both human and rat carcinogenicity, but there are few models to quantitatively assess carcinogenicity in humans. To our knowledge, slope factor (SF), a parameter describing carcinogenicity potential used especially for human risk assessment of contaminated sites, has never been modeled for both inhalation and oral exposures. In this study, we developed classification and regression models for inhalation and oral SFs using data from the Risk Assessment Information System (RAIS) and different machine learning approaches. The models performed well in classification, with accuracies for the external set of 0.76 and 0.74 for oral and inhalation exposure, respectively, and r² values of 0.57 and 0.65 in the regression models for oral and inhalation SFs in external validation. These models might therefore support regulators in (de)prioritizing substances for regulatory action and in weighing evidence in the context of chemical safety assessments. Moreover, these models are implemented on the VEGA platform and are now freely downloadable online.

Purpose This study explored several strategies to improve the performance of literature QSAR models for plasma protein binding (PPB), such as a suitable endpoint transformation, a correct representation of chemicals, more consistency in the dataset, and a reliable definition of the applicability domain. Methods We retrieved human fraction unbound (Fu) data for 670 compounds from the literature and carefully checked them for consistency. Descriptors were calculated taking account of the ionization state of molecules at physiological pH (7.4), in order to better estimate the affinity of molecules to blood proteins. We used different algorithms and chemical descriptors to explore the most suitable strategy for modeling the endpoint. SMILES (simplified molecular input line entry system)-based string descriptors were also tested with the CORAL software (CORelation And Logic). We did an outlier analysis to establish the models to use (or not to use) in case of well recognized families. Results Internal validation of the selected models returned Q² values close to 0.60. External validation also gave r² values always greater than 0.60. The CORAL descriptor based model for √fu was the best, with r² 0.74 in external validation. Conclusions Performance in prediction confirmed the robustness of all the derived models and their suitability for real-life purposes, i.e. screening chemicals for their ADMET profiling. Optimization of descriptors can be useful in order to obtain the correct results with a ionized molecule.