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

An Attention-Driven Graph Transformer With Nonlinear Modeling and Neuro-Fuzzy Fusion for High-Order Toxic Molecular

Phu Pham1

  • 1Faculty of Information Technology, HUTECH University, Ho Chi Minh City, Vietnam.

Molecular Informatics
|May 9, 2026
PubMed
Summary
This summary is machine-generated.

Related Concept Videos

Classification of Neurotransmitters01:30

Classification of Neurotransmitters

Neurotransmitters play a crucial role in the communication between neurons in the autonomic nervous system. Neurons in the autonomic nervous system can be cholinergic or adrenergic depending on the neurotransmitters synthesized. Cholinergic neurons use acetylcholine as their primary neurotransmitter. This includes all the preganglionic fibers of the sympathetic and pre- and postganglionic fibers of the parasympathetic nervous systems. In addition, neurons of the somatic nervous system also use...

You might also read

Related Articles

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

Sort by
Same author

An Integrated Fuzzy Neural Network and Topological Data Analysis for Molecular Graph Representation Learning and Property Forecasting.

Molecular informatics·2025
Same author

A Topology-Enhanced Multi-Viewed Contrastive Approach for Molecular Graph Representation Learning and Classification.

Molecular informatics·2025
Same journal

SpaceExpander: An Automated System for Drafting Markush Claims to Expand Chemical Space.

Molecular informatics·2026
Same journal

A Structure-Informed Atlas of Venom-Derived Peptides Reveals the Organization of Chemical Space.

Molecular informatics·2026
Same journal

ConGen: Targeted Molecule Generation Through Contrastive Learning and Latent Optimization.

Molecular informatics·2026
Same journal

Novel Molecules Generation Using Graph Generative Adversarial Networks.

Molecular informatics·2026
Same journal

Molecular Modeling and Chemoinformatics in Ukraine.

Molecular informatics·2026
Same journal

ADME-DTI: Augmented Deep Meta Ensemble for Drug-Target Interaction Prediction.

Molecular informatics·2026
See all related articles

The novel AKAGTL model enhances toxic molecular graph learning by integrating Kolmogorov-Arnold Networks (KANs) and attention mechanisms. This approach improves regression accuracy for predicting molecular toxicity by better capturing complex interactions and uncertainty.

Area of Science:

  • Computational Chemistry
  • Machine Learning
  • Toxicology

Background:

  • Predicting molecular toxicity is crucial but challenging due to complex, size-varied molecular graphs.
  • Existing deep learning (DL) and graph neural network (GNN) methods struggle with nonlinear expressiveness, high-order structures, and uncertainty.
  • Current graph transformers lack explicit modeling of complex atomic interactions and uncertainty.

Purpose of the Study:

  • To develop an advanced graph transformer framework for toxic molecular graph embedding and regression.
  • To address limitations in nonlinear expressiveness, high-order structural information, and uncertainty handling in current models.
  • To introduce a novel approach that explicitly models complex atomic interactions and uncertainty.

Main Methods:

Keywords:
Kolmogorov–Arnold network (KAN)graph neural network (GNN)graph regressiongraph transformerhigh‐order interactiontoxic molecule

Related Experiment Videos

  • Proposed the Attention-driven Kolmogorov-Arnold Network (KAN)-based Graph Transformer (AKAGTL) model.
  • Incorporated structured KAN-based functional transformations for high-order atomic interactions.
  • Integrated high-order structural representations and a Gaussian neuro-fuzzy fusion mechanism for uncertainty-aware aggregation.

    • Main Results:

    • The AKAGTL model demonstrated consistent improvements in regression accuracy on molecular toxicity benchmarks.
    • Outperformed representative GNN and graph transformer baselines in toxic molecular graph embedding and regression tasks.
    • Showcased the effectiveness of jointly modeling nonlinear interactions, structural dependencies, and uncertainty.

    Conclusions:

    • The AKAGTL model offers a more expressive and robust solution for toxic molecular graph learning.
    • Explicitly modeling nonlinear functional interactions and structural dependencies enhances predictive accuracy.
    • Uncertainty-aware fusion is critical for aggregating heterogeneous feature spaces in molecular property prediction.