Jove
Visualize
Contact Us
JoVE
x logofacebook logolinkedin logoyoutube logo
ABOUT JoVE
OverviewLeadershipBlogJoVE Help Center
AUTHORS
Publishing ProcessEditorial BoardScope & PoliciesPeer ReviewFAQSubmit
LIBRARIANS
TestimonialsSubscriptionsAccessResourcesLibrary Advisory BoardFAQ
RESEARCH
JoVE JournalMethods CollectionsJoVE Encyclopedia of ExperimentsArchive
EDUCATION
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab ManualFaculty Resource CenterFaculty Site
Terms & Conditions of Use
Privacy Policy
Policies

Related Experiment Videos

A comparative study of machine learning algorithms applied to predictive toxicology data mining.

Daniel C Neagu1, Gongde Guo, Paul R Trundle

  • 1Department of Computing, University of Bradford, Bradford, UK. D.Neagu@bradford.ac.uk

Alternatives to Laboratory Animals : ATLA
|April 7, 2007
PubMed
Summary
This summary is machine-generated.

Related Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by
Same author

The Alarming Consequences of Workforce Reductions at the FDA, EPA, NIH and CDC in the United States.

NAM journal·2026
Same author

Development of Acute-to-Chronic Ratios (ACRs) to Support Ecotoxicity Prediction for Surfactants.

Environmental toxicology and chemistry·2026
Same author

MechoA+: A Chemical Structure Profiler Raising the Bar for the Prediction of Mechanisms of Toxic Action for Chemical Safety Assessment.

Environmental science & technology·2026
Same author

Redefining the gold standard and recommendations to accelerate the adoption of New Approach Methodologies (NAMs) in support of the Three Rs.

Alternatives to laboratory animals : ATLA·2026
Same author

The Findable, Accessible, Interoperable, Reusable (FAIR) Lite Principles to ensure utility of computational toxicology models.

ALTEX·2025
Same author

Challenges and opportunities of read-across for the tumor promotion effects of microcystins.

Regulatory toxicology and pharmacology : RTP·2025
Same journal

Spotlight on Three Rs Progress.

Alternatives to laboratory animals : ATLA·2026
Same journal

Resources Round-up.

Alternatives to laboratory animals : ATLA·2026
Same journal

Editorial.

Alternatives to laboratory animals : ATLA·2026
Same journal

<i>Galleria mellonella</i> as an invertebrate screening model for biocompatibility and <i>in vivo</i> functionality testing of imaging contrast agents.

Alternatives to laboratory animals : ATLA·2026
Same journal

Data-driven science, ethics and risk: Building knowledge of key ethics-risk intersections relevant to public sector decision-making.

Alternatives to laboratory animals : ATLA·2026
Same journal

Conference Diary.

Alternatives to laboratory animals : ATLA·2026
See all related articles

This study compared machine learning algorithms for predictive toxicology. Optimal feature selection and model development are crucial for accurate chemical compound classification, not just algorithm choice.

Area of Science:

  • Computational toxicology
  • Cheminformatics
  • Machine learning applications

Background:

  • Predictive toxicology is essential for chemical safety assessment.
  • Machine learning (ML) offers powerful tools for analyzing complex toxicological data.
  • Evaluating diverse ML algorithms on real-world toxicity datasets is critical for understanding their performance.

Purpose of the Study:

  • To comparatively evaluate widely used machine learning algorithms for predictive toxicology data mining.
  • To identify key factors influencing model accuracy in toxicological predictions.
  • To explore the relationship between data descriptors and ML algorithm performance.

Main Methods:

  • Selection of representative and diverse machine learning algorithms.

Related Experiment Videos

  • Extensive evaluation using seven real-world toxicity datasets.
  • Visual analysis of descriptor correlations and their impact on model performance.

    • Main Results:

    • No single machine learning algorithm consistently outperformed others across all seven toxicity datasets.
    • Accurate classification models can often be developed using a limited number of descriptors (up to five).
    • Model performance is sensitive to the chosen feature selection methods and model development techniques.

    Conclusions:

    • The optimal subset of descriptors is as important as algorithm selection for building accurate predictive toxicology models.
    • Overly complex or simplistic descriptor sets can negatively impact model performance.
    • Tailoring feature selection and model development to specific datasets is recommended for effective toxicological predictions.