Computational Toxicology最新文献

An in silico model for predicting skin penetration of active ingredients formulated in plant protection products (PPP) has been developed using random forests (machine learning technique) that were trained with data from in vitro human skin studies taken from the EFSA dermal absorption database and in-house data from Bayer. In addition to the applied dose, various physicochemical properties were considered as model parameters. The model has been linked to a novel percentile approach in order to make the results usable for regulatory purposes. Application to an external validation data set demonstrated that the tool is ready for use. Finally, we propose to follow a tiered decision tree approach for non-dietary risk assessments including the use of the in silico dermal absorption prediction model as part of a safety assessment of a PPP.

The Threshold of Toxicological Concern (TTC) is a pragmatic approach used to establish safe thresholds below which there can be no appreciable risk to human health. Here, a large inventory of ∼45,000 substances (referred to as the LRI dataset) was profiled through the Kroes TTC decision module within Toxtree v3.1 to assign substances into their respective TTC categories. Four thousand and two substances were found to be not applicable for the TTC approach. However, closer examination of these substances uncovered several implementation issues: substances represented in their salt forms were automatically assigned as not appropriate for TTC when many of these contained essential metals as counter ions which would render them TTC applicable. High Potency Carcinogens and dioxin-like substances were not fully captured based on the rules currently implemented in the software. Phosphorus containing substances were considered exclusions when many of them would be appropriate for TTC. Refinements were proposed to address the limitations in the current software implementation. A second component of the study explored a set of substances representative of those released from medical devices and compared them to the LRI dataset as well as other toxicity datasets to investigate their structural similarity. A third component of the study sought to extend the exclusion rules to address application to substances released from medical devices that lack toxicity data. The refined rules were then applied to this dataset and the TTC assignments were compared. This case study demonstrated the importance of evaluating the software implementation of an established TTC workflow, identified certain limitations and explored potential refinements when applying these concepts to medical devices.

Per- and Polyfluoroalkyl substances (PFAS) are a class of synthetic chemicals that are in widespread use and present concerns for persistence, bioaccumulation and toxicity. Whilst a handful of PFAS have been characterised for their hazard profiles, the vast majority of PFAS have not been studied. The US Environmental Protection Agency (EPA) undertook a research project to screen ∼150 PFAS through an array of different in vitro high throughput toxicity and toxicokinetic tests in order to inform chemical category and read-across approaches. A previous publication described the rationale behind the selection of an initial set of 75 PFAS, whereas herein, we describe how various category approaches were applied and extended to inform the selection of a second set of 75 PFAS from our library of approximately 430 commercially procured PFAS. In particular, we focus on the challenges in grouping PFAS for prospective analysis and how we have sought to develop and apply objective structure-based categories to profile the testing library and other PFAS inventories. We additionally illustrate how these categories can be enriched with other information to facilitate read-across inferences once experimental data become available. The availability of flexible, objective, reproducible and chemically intuitive categories to explore PFAS constitutes an important step forward in prioritising PFAS for further testing and assessment.

Estrogen receptor (ER) mediated endocrine disruption and blood–brain barrier (BBB) permeability are two crucial pharmacological endpoints that must be assessed for any drug candidate. However, experimental testing is expensive and time-consuming, and in recent years, Quantitative Structure-Property Relationships (QSPRs) have emerged as a viable in silico alternative. However, most QSPR models reported on ER toxicity and BBB permeability have been carried out using 2D descriptors, whereas it has been established that ER binding and BBB permeability are stereoselective processes in which the spatial arrangement of atoms in the molecule plays a key role. The current study addresses this problem using a chirality-sensitive 3D-QSPR methodology entitled ‘EigenValue ANalysiS (EVANS). The EVANS approach merges information from 3D molecular structure with 2D physicochemical properties to generate eigenvalues which are used as descriptors in QSPR modelling. For chiral compounds, EVANS computes descriptors by considering distance attributes from a plethora of enantiomeric states, thereby accounting for the contributions of multiple conformers towards a particular biological endpoint. We deploy the EVANS methodology with machine learning algorithms to build predictive QSPR models for estrogen receptor (ER) mediated endocrine disruption and BBB permeability. Regression analyses of ER binding on a dataset of 132 chemical entities returned a robust and predictive model, with the support vector machine model having $r_{\mathrm{train}}^2=0.84$ and $r_{\mathrm{test}}^2=0.70$. Classification models for BBB permeability on a dataset of 607 chemicals also showed high prediction accuracy, with the artificial neural network model showing the best performance (Accuracy = 0.85, AUC = 0.82, precision = 0.85, F1 score = 0.89). For comparison, conventional 2D-QSPR models were also built for these endpoints, and it was observed that EVANS generates eigenvalues that are superior to descriptors used in standard 2D-QSPR. Overall, our results demonstrate that EVANS is a powerful 3D-QSPR methodology that offers several advantages over existing QSAR/QSPR methods, and can be a useful computational tool in the pharmacological and toxicological evaluation of new and existing drugs.

Multiple single-walled carbon nanotubes, nestled in tandem as concentric cylinders, constitute the multi-walled carbon nanotubes. Due to their unique physical and chemical characteristics, the multi-walled carbon nanotubes find applications over diverse fields. Investigational studies in the literature reveal toxic nature of multi-walled carbon nanotubes. Hence, it is important to sense and predict their genotoxicity profile for public safety. Deep learning-based toxicity profile prediction, would hasten the research in the alleviation of toxicity in the products build using the multi-walled carbon nanotubes. The proposed hybrid-deep learning framework predicts the genotoxicity of variants of multi-walled carbon nanotubes with higher accuracy and precision. The proposed Ensemble Super Learner (ESL) is a hybrid model, built as a cascade combination of three machine learning models and deep autoencoder. The model achieves cent-percent accuracy when trained over the sparse data available on the genotoxic profile of variants of multi-walled carbon nanotubes.

To address the need for rapid assessment of neurotoxicity from potential exposure to molecules of unknown toxicity, we developed an in silico tool that employs reverse molecular docking to identify receptor targets for molecules and deep-learning models that predict activity on the neurological targets. A selection of human neurologic receptors were obtained from the Protein Data Bank (PDB), then curated and prepared for docking. In total we docked thousands of molecules onto multiples sites on multiple different neurological receptor structures, generating millions of docked poses and scores. With this data we identified protein and ligand interactions and compared that to previously described experimental results. The data was transformed to an image representation and used to generate 2D convolutional deep-learning models. We have generated 19 deep-learning models, of which 17 are over 90% accurate on validation data and the remaining two are 84% and 87% accurate. We have developed a reverse docking GUI and pipeline to identify potential neurological targets for toxins and predict activity of toxins with deep-learning models based on docking identified interactions as an input. As an example, we have applied this pipeline to toluene, a molecule with known toxicity, and correctly predicted it as a GABA(B) agonist. The GUI has been tested on Ubuntu 20.04LTS and Windows 10, and the code, models and GUI are available under GPLv3 on github at https://github.com/mmccarthy1/Autodock_deeplearning_toxicology_GUI.

Bioinformatics is inherently an innovative field that is situated at the limit of life and computer sciences that allowed new technological advances in genome sequencing, data processing, predication and simplified the treatment of complex and huge data. This field is related on two common approaches namely; in silico and molecular docking-dynamic experimentations to improve and clarify the scientific perception of ligand-receptor interactions, especially of those molecules involved in the drug elaboration process. This discipline has emerged to replace the traditional approach of drug discovery which was very limited, very expensive, and didn’t always provide the expected results. The objective of this review is to report the key events that have marked the bioinformatics sector during these last few years but also to underline the key elements that have contributed to its success especially in the sectors of pharmacy, biotechnology, bioengineering, and teaching but also on scientific community cooperation. This review will also discuss cutting-edge technology and bioinformatics characteristics in order to clarify some ambiguities in this area.

Context

Relationships among asbestos exposure, lung burden, and mesothelioma risks have been previously evaluated, but it would be useful to validate published epidemiological observations with a mathematical model describing deposition and elimination of various mineral types of fibers.

Objective

(a) To develop a mechanistical model demonstrating uptake and removal of fibers from human lungs, (b) To test the model on the results of a British case-control study, (c) To quantify the updated values for elimination coefficient of various mineral types of asbestos fibers.

Methods

A mechanistic model utilizing the first-order kinetic relationship is proposed that relates levels of exposure to mineral fibers, elimination coefficients, and lung burden at certain points of time. The behaviour of the model was explored for different exposure scenarios. Elimination coefficients for various mineral types were estimated based on the observed proportion of asbestos minerals in exposure vs observed lung burden.

Results

Based on the proposed model, the average elimination coefficient was estimated for crocidolite as 0.099 vs average published value of 0.092, for amosite as 0.169 vs 0.19, and for chrysotile as 6.45 vs average published value of 6.36 (years⁻¹). Lung burden level was demonstrated to change linearly with exposure intensity, and supra-linearly with exposure duration. The simulation of three separate exposure events during three decades showed that lung burden level prevailingly depends on the most recent event (R = 0.967, p < 0.05) and only weakly correlates with the most remote event (R = 0.032, p < 0.05).

Conclusion

In spite of potential limitations, mechanistical modelling of asbestos exposure can serve as an effective tool for risk assessment purposes.

The Predictive Analytics Toolkit (PAT) was developed to facilitate use of new approach methodologies (NAMs) to predict health hazards and risks from chemicals. PAT is a user-friendly web application that integrates many R packages to enable development and testing of prediction models without any programming. We drew from the work of Ring et al. 2021 (https://doi.org/10.1016/j.comtox.2021.100166), who used random forest models to predict in vivo transcriptomic responses in rat liver from in vitro Tox21 AC50 values for a set of 221 chemicals. Gene ontologies helped identify 735 biological pathways based on differential in vivo expression of specific gene sets. Ring et al. used 12 models that varied in use of toxicokinetics to predict in vivo activity using 5000 random forest iterations for each chemical/pathway combination (the area under the receiver-operator characteristic curve (AUC-ROC) was the measure of model performance). The highest-ranking model (Model 10) used Tox21 AC50 nominal concentrations converted to media concentrations and in vivo doses converted to circulating plasma concentrations; the lowest ranking model (Model 2) used nominal in vitro concentrations and administered in vivo dose levels. Using a subset of 10 pathways from the Ring et al. data, we used PAT to predict the AUC-ROC and to compare the best (Model 10) and worst (Model 2) performing models with only 100 random forest iterations. Using the results from PAT, Model 10 “won” in 60% of the comparisons, a value similar to that calculated for the identical set of comparisons using the supplemental data from Ring et al. (52.2%). Hence, PAT can provide a useful alternative to programming in R for prediction modeling and model performance evaluation, even for extensive genomic data sets.

Mucormycosis or “black fungus” has been currently observed in India, as a secondary infection in COVID-19 infected patients in the post-COVID-stage. Fungus is an uncommon opportunistic infection that affects people who have a weak immune system. In this study, 158 antifungal phytochemicals were screened using molecular docking against glucoamylase enzyme of Rhizopus oryzae to identify potential inhibitors. The docking scores of the selected phytochemicals were compared with Isomaltotriose as a positive control. Most of the compounds showed lower binding energy values than Isomaltotriose (-6.4 kcal/mol). Computational studies also revealed the strongest binding affinity of the screened phytochemicals was Dioscin (-9.4 kcal/mol). Furthermore, the binding interactions of the top ten potential phytochemicals were elucidated and further analyzed. In-silico ADME and toxicity prediction were also evaluated using SwissADME and admetSAR online servers. Compounds Piscisoflavone C, 8-O-methylaverufin and Punicalagin exhibited positive results with the Lipinski filter and drug-likeness and showed mild to moderate of toxicity. Molecular dynamics (MD) simulation (at 300 K for 100 ns) was also employed to the docked ligand-target complex to explore the stability of ligand-target complex, improve docking results, and analyze the molecular mechanisms of protein-target interactions.