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

Toward an optimal procedure for variable selection and QSAR model building.

A Yasri1, D Hartsough

  • 1Computational Design Group, ArQule Inc., 19 Presidential Way, Woburn, MA 01801, USA. ayasri@arqule.com

Journal of Chemical Information and Computer Sciences
|October 18, 2001
PubMed
Summary

This study introduces a novel Quantitative Structure-Activity Relationship (QSAR) method using genetic algorithms and neural networks. It efficiently selects relevant molecular descriptors and optimizes QSAR models, especially for complex, nonlinear data.

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

Recent advances in drug design of epidermal growth factor receptor inhibitors.

Current medicinal chemistry·2013
Same author

Computer-assisted rational design of immunosuppressive compounds.

Nature biotechnology·1998
Same author

17 alpha (haloacetamidoalkyl) estradiols alkylate the human estrogen receptor at cysteine residues 417 and 530.

Biochemistry·1997
Same author

Rational choice of molecular dynamics simulation parameters through the use of the three-dimensional autocorrelation method: application to calmodulin flexibility study.

Protein engineering·1996

Area of Science:

  • Computational Chemistry
  • Cheminformatics
  • Machine Learning in Drug Discovery

Background:

  • Quantitative Structure-Activity Relationship (QSAR) studies are crucial for predicting biological activity.
  • Traditional QSAR methods often struggle with high-dimensional descriptor spaces and complex nonlinear relationships.
  • Efficient variable selection and model architecture optimization are key challenges in QSAR.

Purpose of the Study:

  • To develop a novel QSAR technique integrating genetic algorithms (GAs) and neural networks (NNs).
  • To enable simultaneous selection of relevant molecular descriptors and optimization of NN architecture.
  • To improve the performance of QSAR models, particularly for nonlinear datasets.

Main Methods:

  • A hybrid approach combining GAs for descriptor selection and NNs for property mapping.

Related Experiment Videos

  • GA search is unconstrained by a fixed number of descriptors.
  • NN architecture, including hidden layer size, is dynamically optimized in parallel with descriptor selection.

    • Main Results:

    • The developed QSAR technique successfully builds both classification and regression models.
    • Demonstrated superior performance compared to simpler variable selection methods on nonlinear datasets.
    • Validated using both artificial and real biological data, showing competitive results against existing techniques.

    Conclusions:

    • The novel GA-NN QSAR technique offers an effective strategy for building robust predictive models.
    • Dynamic optimization of both descriptors and model architecture enhances QSAR model accuracy.
    • Highlights best practices and important considerations for developing reliable QSAR models.