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

SAR modeling of unbalanced data sets.

H S Rosenkranz1, A R Cunningham

  • 1Department of Environmental and Occupational Health, Graduate School of Public Health, University of Pittsburgh, 111 Parran Hall, 130 DeSoto Street, Pittsburgh, PA 15261, USA.

SAR and QSAR in Environmental Research
|November 8, 2001
PubMed
Summary

Structure-Activity Relationship (SAR) models for hazard identification perform adequately with unbalanced active-to-inactive chemical ratios between 3:1 and 1:2. Ratios exceeding 4:1 lead to unsatisfactory predictive performance, even with corrections.

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

Describing county-level clustering patterns of suicide among Veterans Health Administration patients and all American adults, 2018-2019, using exploratory spatial data analysis.

Annals of epidemiology·2025
Same author

The Development and Characterisation of a Structure-activity Relationship Model of the Draize Eye Irritation Test.

Alternatives to laboratory animals : ATLA·2015
Same author

The High Production Volume Chemical Challenge Program : The Rodent LD50 and its Possible Replacement.

Alternatives to laboratory animals : ATLA·2014
Same author

Structural Basis of the Toxicity of Chemicals in Cultured Human HeLa Cells.

Alternatives to laboratory animals : ATLA·2014
Same author

A battery of cell toxicity assays as predictors of eye irritation: a feasibility study.

Alternatives to laboratory animals : ATLA·2014
Same author

Structure-activity relationship models for rat carcinogenesis and assessing the role mutagens play in model predictivity.

SAR and QSAR in environmental research·2014

Area of Science:

  • Quantitative Structure-Activity Relationship (QSAR) modeling
  • Toxicology and Hazard Identification
  • Computational Chemistry

Background:

  • Structure-Activity Relationship (SAR) approaches are increasingly used for chemical hazard identification.
  • Optimizing the predictive performance of SAR models is crucial for reliable hazard assessment.
  • The composition of the learning set, specifically the ratio of active to inactive chemicals, can significantly impact model performance.

Purpose of the Study:

  • To investigate the impact of the active-to-inactive chemical ratio in learning sets on SAR model predictive performance.
  • To determine the acceptable range of this ratio for adequate model performance.
  • To identify thresholds beyond which SAR model performance becomes unsatisfactory.

Main Methods:

Related Experiment Videos

  • Development and evaluation of SAR models using datasets with varying ratios of active to inactive chemicals.
  • Statistical analysis to assess model performance across different learning set compositions.
  • Correction methods for initial ratio imbalances were applied and evaluated.

Main Results:

  • SAR models demonstrated adequate performance with active-to-inactive ratios ranging from 3:1 (75% active) to 1:2 (33% active).
  • Theoretical optimal ratio is 1:1, but models tolerate significant imbalance.
  • Models trained on learning sets with ratios exceeding 4:1 (80% active) performed unsatisfactorily, even after ratio correction.

Conclusions:

  • The ratio of active to inactive compounds in a SAR learning set significantly influences model predictive performance.
  • A balanced ratio is not strictly necessary, with acceptable performance observed within a defined range of imbalance.
  • Careful consideration of learning set composition is essential for developing robust and reliable SAR models for hazard identification.