Leveraging high-throughput screening data, deep neural networks, and conditional generative adversarial networks to advance predictive toxicology

Green, Adrian; Mohlenkamp, Martin J.; Das, Jhuma; Chaudhari, Meenal; Truong, Lisa; Tanguay, Robert L.; Reif, David M.

doi:10.1371/journal.pcbi.1009135

Cited by 25 publications

(19 citation statements)

References 49 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…For all models, the performance on the external test set was slightly better than the internal performance. Although the diversity was ensured between all molecules of the training and the external test sets with a maximal chemical similarity of 0.85, that may happen due to the random choice of the molecules for the external set, that was also observed in other recent ML modeling studies [ 39 , 40 ]. The excellent performance obtained with the 36 best MOE and 7 IE descriptors suggest that these RF and SVM models can be employed to find new CYP2C9 inhibitors.…”

Section: Resultsmentioning

confidence: 92%

Machine learning-driven identification of drugs inhibiting cytochrome P450 2C9

et al. 2022

View full text Add to dashboard Cite

Cytochrome P450 2C9 (CYP2C9) is a major drug-metabolizing enzyme that represents 20% of the hepatic CYPs and is responsible for the metabolism of 15% of drugs. A general concern in drug discovery is to avoid the inhibition of CYP leading to toxic drug accumulation and adverse drug–drug interactions. However, the prediction of CYP inhibition remains challenging due to its complexity. We developed an original machine learning approach for the prediction of drug-like molecules inhibiting CYP2C9. We created new predictive models by integrating CYP2C9 protein structure and dynamics knowledge, an original selection of physicochemical properties of CYP2C9 inhibitors, and machine learning modeling. We tested the machine learning models on publicly available data and demonstrated that our models successfully predicted CYP2C9 inhibitors with an accuracy, sensitivity and specificity of approximately 80%. We experimentally validated the developed approach and provided the first identification of the drugs vatalanib, piriqualone, ticagrelor and cloperidone as strong inhibitors of CYP2C9 with IC values <18 μM and sertindole, asapiprant, duvelisib and dasatinib as moderate inhibitors with IC50 values between 40 and 85 μM. Vatalanib was identified as the strongest inhibitor with an IC50 value of 0.067 μM. Metabolism assays allowed the characterization of specific metabolites of abemaciclib, cloperidone, vatalanib and tarafenacin produced by CYP2C9. The obtained results demonstrate that such a strategy could improve the prediction of drug-drug interactions in clinical practice and could be utilized to prioritize drug candidates in drug discovery pipelines.

show abstract

Section: Resultsmentioning

confidence: 92%

Machine learning-driven identification of drugs inhibiting cytochrome P450 2C9

et al. 2022

View full text Add to dashboard Cite

show abstract

“…In terms of human exposure pattern modeling, a recent notable study analyzed chemical biomonitoring data within the National Health and Nutrition Examination Survey (NHANES) through frequent itemset mining, which yielded novel chemical combinations detected to frequently co-occur in humans [ 42 ]. Other advanced in silico methods have included additional machine-learning approaches (i.e., deep learning) [ 43 ], Bayesian statistical methods [ 44 ], and text mining [ 45 ]. The current study, as well as these additional examples of silico/database-driven approaches, collectively represent progress in the fields of exposure science and toxicology towards better characterizing environmental mixtures.…”

Section: Discussionmentioning

confidence: 99%

Chemical Mixtures in Household Environments: In Silico Predictions and In Vitro Testing of Potential Joint Action on PPARγ in Human Liver Cells

Carberry

Turla

Koval

et al. 2022

Toxics

View full text Add to dashboard Cite

There are thousands of chemicals that humans can be exposed to in their everyday environments, the majority of which are currently understudied and lack substantial testing for potential exposure and toxicity. This study aimed to implement in silico methods to characterize the chemicals that co-occur across chemical and product uses in our everyday household environments that also target a common molecular mediator, thus representing understudied mixtures that may exacerbate toxicity in humans. To detail, the Chemical and Products Database (CPDat) was queried to identify which chemicals co-occur across common exposure sources. Chemicals were preselected to include those that target an important mediator of cell health and toxicity, the peroxisome proliferator activated receptor gamma (PPARγ), in liver cells that were identified through query of the ToxCast/Tox21 database. These co-occurring chemicals were thus hypothesized to exert potential joint effects on PPARγ. To test this hypothesis, five commonly co-occurring chemicals (namely, benzyl cinnamate, butyl paraben, decanoic acid, eugenol, and sodium dodecyl sulfate) were tested individually and in combination for changes in the expression of PPARγ and its downstream target, insulin receptor (INSR), in human liver HepG2 cells. Results showed that these likely co-occurring chemicals in household environments increased both PPARγ and INSR expression more significantly when the exposures occurred as mixtures vs. as individual chemicals. Future studies will evaluate such chemical combinations across more doses, allowing for further quantification of the types of joint action while leveraging this method of chemical combination prioritization. This study demonstrates the utility of in silico-based methods to identify chemicals that co-occur in the environment for mixtures toxicity testing and highlights relationships between understudied chemicals and changes in PPARγ-associated signaling.

show abstract

“…Dr. David Reif is a Professor in the Department of Biological Sciences at North Carolina State University (NCSU) and Director of the NCSU Bioinformatics Consulting and Services Core. Dr. Reif leads studies implementing computational modeling approaches to leverage big data in predicting exposure and disease outcomes ( Kosnik and Reif, 2019 ; Green et al, 2021 ; Marvel et al, 2021 ). Dr. Jaspers is a Professor in the Department of Pediatrics, Microbiology and Immunology at UNC, and is the Director of the Curriculum in Toxicology and Environmental Medicine and Director of the Center for Center for Environmental Medicine, Asthma and Lung Biology.…”

Section: Methodsmentioning

confidence: 99%

Development of the InTelligence And Machine LEarning (TAME) Toolkit for Introductory Data Science, Chemical-Biological Analyses, Predictive Modeling, and Database Mining for Environmental Health Research

et al. 2022

Self Cite

View full text Add to dashboard Cite

Research in environmental health is becoming increasingly reliant upon data science and computational methods that can more efficiently extract information from complex datasets. Data science and computational methods can be leveraged to better identify relationships between exposures to stressors in the environment and human disease outcomes, representing critical information needed to protect and improve global public health. Still, there remains a critical gap surrounding the training of researchers on these in silico methods. We aimed to address this gap by developing the inTelligence And Machine lEarning (TAME) Toolkit, promoting trainee-driven data generation, management, and analysis methods to “TAME” data in environmental health studies. Training modules were developed to provide applications-driven examples of data organization and analysis methods that can be used to address environmental health questions. Target audiences for these modules include students, post-baccalaureate and post-doctorate trainees, and professionals that are interested in expanding their skillset to include recent advances in data analysis methods relevant to environmental health, toxicology, exposure science, epidemiology, and bioinformatics/cheminformatics. Modules were developed by study coauthors using annotated script and were organized into three chapters within a GitHub Bookdown site. The first chapter of modules focuses on introductory data science, which includes the following topics: setting up R/RStudio and coding in the R environment; data organization basics; finding and visualizing data trends; high-dimensional data visualizations; and Findability, Accessibility, Interoperability, and Reusability (FAIR) data management practices. The second chapter of modules incorporates chemical-biological analyses and predictive modeling, spanning the following methods: dose-response modeling; machine learning and predictive modeling; mixtures analyses; -omics analyses; toxicokinetic modeling; and read-across toxicity predictions. The last chapter of modules was organized to provide examples on environmental health database mining and integration, including chemical exposure, health outcome, and environmental justice indicators. Training modules and associated data are publicly available online (https://uncsrp.github.io/Data-Analysis-Training-Modules/). Together, this resource provides unique opportunities to obtain introductory-level training on current data analysis methods applicable to 21st century science and environmental health.

show abstract

Leveraging high-throughput screening data, deep neural networks, and conditional generative adversarial networks to advance predictive toxicology

Cited by 25 publications

References 49 publications

Machine learning-driven identification of drugs inhibiting cytochrome P450 2C9

Machine learning-driven identification of drugs inhibiting cytochrome P450 2C9

Chemical Mixtures in Household Environments: In Silico Predictions and In Vitro Testing of Potential Joint Action on PPARγ in Human Liver Cells

Development of the InTelligence And Machine LEarning (TAME) Toolkit for Introductory Data Science, Chemical-Biological Analyses, Predictive Modeling, and Database Mining for Environmental Health Research

Contact Info

Product

Resources

About