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

Predicting activities without computing descriptors: graph machines for QSAR.

A Goulon1, T Picot, A Duprat

  • 1Laboratoire d'Electronique, Ecole Supérieure de Physique et de Chimie Industrielles de la Ville de Paris (ESPCI-ParisTech), 10 rue Vauquelin, 75005 Paris, France.

SAR and QSAR in Environmental Research
|March 17, 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

Real-world EGFR testing practices for non-small-cell lung cancer by thoracic pathology laboratories across Europe.

ESMO open·2023
Same author

[Multiple synchronous lung adenocarcinoma: Interest of molecular biology for staging].

Revue des maladies respiratoires·2021
Same author

Herbage intake and growth of rabbits under different pasture type, herbage allowance and quality conditions in organic production.

Animal : an international journal of animal bioscience·2018
Same author

Biomechanical study comparing Pulvertaft suture to step-cut suture.

Hand surgery & rehabilitation·2017
Same author

PASTRAB: a model for simulating intake regulation and growth of rabbits raised on pastures.

Animal : an international journal of animal bioscience·2017
Same author

Surgery of acute type A dissection: what have we learned during the past 25 years?

Zeitschrift fur Kardiologie·2016
Same journal

Mapping toxicity pathways of per- and polyfluoroalkyl substances using interpretable classification-based machine learning models.

SAR and QSAR in environmental research·2026
Same journal

Structure-based identification of inhibitory compounds targeting M32 metallocarboxypeptidase of <i>Leishmania donovani</i>.

SAR and QSAR in environmental research·2026
Same journal

Multiscale computational evaluation of marine fungal metabolites containing iminohydantoin-like scaffolds as anti-Alzheimer drug candidates.

SAR and QSAR in environmental research·2026
Same journal

Conformational landscapes and binding free energies of multitarget phytochemicals reveal molecular recognition mechanisms in colorectal cancer-associated proteins.

SAR and QSAR in environmental research·2026
Same journal

AI-driven QSAR modelling and virtual screening in the discovery of selective dopamine D<sub>2</sub> receptor ligands.

SAR and QSAR in environmental research·2026
Same journal

Integrating machine learning and pharmacogenomics for biomarker discovery, identification and prioritization of potential drug candidates in ovarian cancer.

SAR and QSAR in environmental research·2026
See all related articles

Graph machines offer a novel approach to quantitative structure-activity relationship (QSAR) modeling by representing molecules as graphs. This method bypasses the need for traditional molecular descriptors, improving efficiency and performance in chemical activity prediction.

Area of Science:

  • Computational Chemistry
  • Machine Learning
  • Cheminformatics

Background:

  • Traditional quantitative structure-activity relationship (QSAR) and quantitative structure-property relationship (QSPR) models often rely on complex molecular descriptors.
  • Designing, computing, and selecting optimal descriptors can be a significant challenge in cheminformatics.

Purpose of the Study:

  • To introduce graph machines as an alternative to traditional machine learning-based QSAR methods.
  • To demonstrate that graph machines can circumvent the challenges associated with molecular descriptor selection and computation.

Main Methods:

  • Molecules are represented as graphs, treating them as structured data.
  • A mathematical function, termed a 'graph machine,' is constructed for each data set example, mirroring the molecule's structure.

Related Experiment Videos

  • Identical parameterized functions ('node functions,' e.g., feedforward neural networks) are combined, with shared weights adjusted during training.
  • Model selection is performed using standard cross-validation techniques.

    • Main Results:

    • Graph machines were successfully applied to various QSAR/QSPR tasks and modeling complex chemical activities.
    • The approach demonstrated superior performance compared to traditional methods in predicting activities like phenol toxicity and anti-HIV activity of HEPT derivatives.
    • The method effectively bypasses the need for manual descriptor selection and computation.

    Conclusions:

    • Graph machines provide an efficient and effective alternative to descriptor-based QSAR/QSPR modeling.
    • This approach simplifies the modeling process by directly utilizing molecular structure.
    • The method shows significant promise for predicting chemical activities and properties.