Physiological variables in machine learning QSARs allow for both cross-chemical and cross-species predictions

Zubrod, Jochen P.; Galić, Nika; Vaugeois, Maxime; Dreier, David A.

doi:10.1016/j.ecoenv.2023.115250

Cited by 7 publications

(16 citation statements)

References 58 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Similar to us, Zubrod et al . (2023) modeled multiple species simultaneously and included species-specific data from the Add my Pet database in addition to chemical properties and the molecular representations ToxPrint and Mordred to predict log10-transformed mass LC50 7 . However, their chemical space was limited to pesticides.…”

Section: Resultsmentioning

confidence: 99%

“…Zubrod et al . (2023) referred to models also including species-specific and experimental information as Bio-QSARs 7 . ML methods can be applied to both QSARs and extended QSARs with non-chemical features.…”

Section: Introductionmentioning

confidence: 99%

“…Besides chemical properties and the molecular representations and Mordred, Zubrod et al . (2023) included species-specific data from the Add my Pet database 7 . They trained RF models on freshwater fish and algae datasets.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

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

Section: Resultsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

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

show abstract

“…These methods either follow the one chemical–one species–one outcome approach, extrapolate the effects of multiple chemicals to one species (e.g., using quantitative structure activity relationships, QSARs), or can extrapolate effects of a given chemical on multiple species. Cross-species extrapolation research has been growing with approaches developed using linear regression models, read-across-based methods, or machine learning (ML). , Through these extrapolations, data gaps can be filled so that chemical effects on a variety of species can be better characterized. For example, Sequence Alignment to Predict Across Species Susceptibility (SeqAPASS) is an online tool that uses a tiered approach to classify species similarity, starting from primary amino acid sequences through to conserved functional domains and likely protein conformations/interactions with chemical stressors .…”

Section: The Need For Computational Methods To Assess Chemical Impact...mentioning

confidence: 99%

“…However, rigorous testing and validation of ML models with benchmark data are crucial for the advancement of these methods . Other species-specific parameters like ecological trait data, TKTD parameters, and/or dynamic energy budget parameters (e.g., the Add-my-Pet database ) can also be used to improve ML-based extrapolations, thus reducing the need for additional animal testing. ,, Further, as other approaches improve (e.g., identification of common protein markers across species or AOPs), so too can ML methods become more robust and reliable.…”

Section: The Need For Computational Methods To Assess Chemical Impact...mentioning

confidence: 99%

Harnessing Computational Methods to Characterize Chemical Impacts on Biodiversity

Kosnik,

Schuwirth,

Rico

2024

Environ. Sci. Technol. Lett.

View full text Add to dashboard Cite

Chemical pollution can threaten biodiversity at different levels, from genetically diverse populations (genetic diversity) to different species (species diversity) and ecosystem traits/interactions (functional diversity). Most assessments of chemical impacts on different biodiversity levels depend on wet lab and field experiments, including sequencing large numbers of organisms, environmental DNA approaches, single chemical−species− outcome toxicity tests, and trait-based methods. However, it is impossible to assess all chemicals, species, populations, and ecosystems using these methods. Therefore, we advocate that computational methods are necessary to characterize, quantify, and predict chemical impacts on biodiversity. We briefly introduce the current state of research into chemical impacts on genetic diversity, species diversity, and functional diversity and describe new opportunities for computational methods like data integration, machine learning, cross-species/cross-ecosystem extrapolation, adverse outcome pathways, and Bayesian methods to support research in these three areas. By harnessing data and methods currently at our disposal and preparing methods to take advantage of continuously emerging data sets, computational approaches can be paired with environmental monitoring so different levels of biological organization can serve as consecutive warning signs for chemical impacts on biodiversity. This will enable effective ecosystem protection measures to be better developed and implemented to prevent biodiversity loss from chemical pollution.

show abstract

Bio-QSARs 2.0: Unlocking a new level of predictive power for machine learning-based ecotoxicity predictions by exploiting chemical and biological information

Zubrod,

Galic,

Vaugeois

et al. 2024

Environment International

View full text Add to dashboard Cite

Physiological variables in machine learning QSARs allow for both cross-chemical and cross-species predictions

Cited by 7 publications

References 58 publications

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

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

Harnessing Computational Methods to Characterize Chemical Impacts on Biodiversity

Bio-QSARs 2.0: Unlocking a new level of predictive power for machine learning-based ecotoxicity predictions by exploiting chemical and biological information

Contact Info

Product

Resources

About