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

Supervised self-organizing maps in drug discovery. 1. Robust behavior with overdetermined data sets.

Yun-De Xiao1, Aaron Clauset, Rebecca Harris

  • 1Molecular Design Group, Targacept Inc., 200 East First Street, Suite 300, Winston-Salem, North Carolina 27101-4165, USA.

Journal of Chemical Information and Modeling
|November 29, 2005
PubMed
Summary

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

Editorial and peer review dynamics at elite general science journals.

Science advances·2026
Same author

Economic Evaluation of the RETURN Intervention to Increase Routine Dental Attendance Among Disadvantaged Adults: Net Health Benefit (QALYs) and Patient Cost Outcomes From a Randomised Controlled Trial.

Community dentistry and oral epidemiology·2026
Same author

Cross-assay RNA modeling reveals cancer biomarkers.

bioRxiv : the preprint server for biology·2026
Same author

On the predictability of progression-free survival in ovarian cancer from NanoString gene expression data.

bioRxiv : the preprint server for biology·2026
Same author

Positive predictive value of respiratory infection diagnoses in the Danish National Patient Registry: A post hoc validation analysis of the DANFLU-1 trial.

International journal of infectious diseases : IJID : official publication of the International Society for Infectious Diseases·2026
Same author

Prevalence of liver fibrosis in the general population (the LiverScreen project): a multinational European cohort study.

Lancet (London, England)·2026

Supervised Kohonen self-organizing maps (sSOMs) offer superior predictive accuracy in quantitative structure-activity relationship (QSAR) analysis compared to standard methods. sSOMs effectively handle noise and redundancy, improving classification of chemical compounds.

Area of Science:

  • Computational chemistry
  • Cheminformatics
  • Machine learning in drug discovery

Background:

  • Quantitative Structure-Activity Relationship (QSAR) analysis is crucial for predicting drug properties.
  • Traditional QSAR methods often struggle with noisy data and feature redundancy.
  • Self-organizing maps (SOMs) offer nonlinear, topology-preserving mapping for data visualization and clustering.

Purpose of the Study:

  • To evaluate the utility of supervised Kohonen self-organizing maps (sSOMs) in QSAR analysis.
  • To compare the performance of sSOMs against established statistical and machine learning methods.
  • To assess the impact of sSOM training parameters on predictive accuracy.

Main Methods:

  • Utilized supervised Kohonen self-organizing maps (sSOMs) for QSAR modeling.

Related Experiment Videos

  • Compared sSOMs with standard methods: partial least squares, stepwise multiple linear regression, genetic functional algorithm, and genetic partial least squares.
  • Included k-nearest neighbor (kNN) classification for direct comparison.
  • Employed the dihydrofolate reductase (DHFR) inhibition dataset for evaluation.
  • Investigated two sSOM training strategies concerning neighborhood kernel reduction.

    • Main Results:

    • sSOMs demonstrated greater robustness to noise and feature redundancy than most chemometric methods.
    • sSOMs effectively utilized descriptors with only nominal linear correlation to the target property.
    • The sSOM approach yielded more accurate predictions compared to standard linear QSAR methods.
    • Performance was evaluated across varying class resolutions for DHFR inhibition data.

    Conclusions:

    • Supervised Kohonen self-organizing maps represent a powerful nonlinear approach for QSAR analysis.
    • sSOMs provide enhanced predictive accuracy and better handling of complex datasets.
    • This method offers a valuable alternative to traditional linear QSAR techniques.