Making in silico predictive models for toxicology FAIR

Cronin, Mark; Belfield, Samuel J.; Briggs, Katharine; Enoch, Steven J.; Firman, James W.; Frericks, Markus; Garrard, C.J.O.; Maccallum, Peter H; Madden, Judith C.; Pastor, Manuel; Sanz, Ferrán; Soininen, Inari; Sousoni, Despoina

doi:10.1016/j.yrtph.2023.105385

Cited by 10 publications

(13 citation statements)

References 31 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Also, results from toxicity experiments presented in scientific papers are generally not submitted to data repositories and need to be manually extracted from the texts. Introducing modern FAIR data-sharing principles in both academy, government, and industry would, thus, greatly facilitate the development of data-driven approaches within ecotoxicology ( 44 , 45 ).…”

Section: Discussionmentioning

confidence: 99%

Transformers enable accurate prediction of acute and chronic chemical toxicity in aquatic organisms

Gustavsson,

Käll,

Svedberg

et al. 2024

Sci. Adv.

View full text Add to dashboard Cite

Environmental hazard assessments are reliant on toxicity data that cover multiple organism groups. Generating experimental toxicity data is, however, resource-intensive and time-consuming. Computational methods are fast and cost-efficient alternatives, but the low accuracy and narrow applicability domains have made their adaptation slow. Here, we present a AI-based model for predicting chemical toxicity. The model uses transformers to capture toxicity-specific features directly from the chemical structures and deep neural networks to predict effect concentrations. The model showed high predictive performance for all tested organism groups—algae, aquatic invertebrates and fish—and has, in comparison to commonly used QSAR methods, a larger applicability domain and a considerably lower error. When the model was trained on data with multiple effect concentrations (EC 50 /EC 10 ), the performance was further improved. We conclude that deep learning and transformers have the potential to markedly advance computational prediction of chemical toxicity.

show abstract

Section: Discussionmentioning

confidence: 99%

Transformers enable accurate prediction of acute and chronic chemical toxicity in aquatic organisms

Gustavsson,

Käll,

Svedberg

et al. 2024

Sci. Adv.

View full text Add to dashboard Cite

show abstract

“…One of the primary objectives of the present work was to develop simple, interpretable, reproducible, transferable, and FAIR (Findable, Accessible, Interoperable, and Reusable) models. 54 Moreover, simple and linear models have greater acceptability to regulatory agencies like the United States Environmental Protection Agency (US-EPA), European Chemicals Agency (ECHA), etc. Although machine learning models are increasingly being used for prediction purposes, their simplicity, reproducibility, and interpretability may remain an issue to some extent, at least in the regulatory acceptability context.…”

Section: Methodsmentioning

confidence: 99%

Read-across-based intelligent learning: development of a global q-RASAR model for the efficient quantitative predictions of skin sensitization potential of diverse organic chemicals

Banerjee,

Roy

2023

Environ. Sci.: Processes Impacts

View full text Add to dashboard Cite

show abstract

“…In the field of toxicity, these models are sometimes referred to as quantitative structure toxicity relationship (QSTR) models, which have a long history of predicting ecotoxicological outcomes using either linear or nonlinear relationships between chemical descriptors and a biological response 5 . According to an estimate published in 2023, there are more than 10,000 models currently published or publicly available 6 . The field of QSAR research recently has seen the adoption of machine learning (ML), i.e., computational methods that are able to find hidden patterns in large amounts of data without explicit programming and, on the basis of said patterns, are able to make predictions.…”

Section: Introductionmentioning

confidence: 99%

“…For QSAR models, similar quality standards have already been proposed (with 49 assessment criteria covering various aspects of QSAR development, documentation and use) 17 and further developed specifically for the application of ML methods to QSARs 18 . Furthermore, the FAIR (Findable, Accessible, Interoperable, Reusable) principles, which were developed for data sharing, could be adapted to model description and deployment and therefore help to improve the reproducibility and largescale adoption of these methods, and eventually turn them into a (re)usable resource for chemical safety assessment 6 . Data handling, i.e., curation, processing, and use in a modeling framework, plays an equally crucial role to avoid reproducibility issues.…”

Section: Introductionmentioning

confidence: 99%

Machine learning-based prediction of fish acute mortality: Implementation, interpretation, and regulatory relevance

Gasser,

Schür,

Perez-Cruz

et al. 2024

Preprint

View full text Add to dashboard Cite

Regulation of chemicals requires knowledge of their toxicological effects on a large number of target species. Traditionally, this knowledge has been acquired throughin vivotesting. The recent effort to find alternatives based on machine learning, however, has not focused on guaranteeing transparency, comparability and reproducibility, which makes it difficult to assess advantages and disadvantages of these methods. Also, comparable baseline performances are needed. In this study, we trained regression models on the ADORE “t-F2F” challenge proposed in [Schüret al.,Nature Scientific data, 2023] to predict acute mortality, measured as LC50 (lethal concentration 50), of organic compounds on fishes. We trained LASSO, random forest (RF), XGBoost, Gaussian process (GP) regression models, and found a series of aspects that are stable across models: (i) using mass or molar concentrations does not affect performances; (ii) the performances are only weakly dependent on how a chemical is represented, but (iii) strongly on how the data is split. Overall, the tree-based models RF and XGBoost performed best and we were able to predict the log10-transformed LC50 with a root mean square error of 0.90, which corresponds to an order of magnitude on the original LC50 scale. On a local level, on the other hand, the models are not able to predict the toxicity of single chemicals accurately enough. Predictions for single chemicals are dominated by a few chemical properties and taxonomic traits are not captured sufficiently by the models. As a consequence, the models are not yet applicable in the regulatory process. Nevertheless, this work contributes to the ongoing discussions on how machine learning can be integrated in the regulatory process.Environmental significanceConventional environmental hazard assessment in its current form will not be able to adapt to the growing need for toxicity testing. Alternative methods, such as toxicity prediction through machine learning, could fulfill that need in an economically and ethically sound manner. Proper implementation, documentation, and the integration into the regulatory process are prerequisites for the usability and acceptance of these models.

show abstract

Making in silico predictive models for toxicology FAIR

Cited by 10 publications

References 31 publications

Transformers enable accurate prediction of acute and chronic chemical toxicity in aquatic organisms

Transformers enable accurate prediction of acute and chronic chemical toxicity in aquatic organisms

Read-across-based intelligent learning: development of a global q-RASAR model for the efficient quantitative predictions of skin sensitization potential of diverse organic chemicals

Machine learning-based prediction of fish acute mortality: Implementation, interpretation, and regulatory relevance

Contact Info

Product

Resources

About