Search research articles

ABOUT JoVE

Overview Leadership Blog JoVE Help Center

AUTHORS

Publishing Process Editorial Board Scope & Policies Peer Review FAQ Submit

LIBRARIANS

Testimonials Subscriptions Access Resources Library Advisory Board FAQ

RESEARCH

JoVE Journal Methods Collections JoVE Encyclopedia of Experiments Archive

EDUCATION

JoVE Core JoVE Business JoVE Science Education JoVE Lab Manual Faculty Resource Center Faculty Site

Terms & Conditions of Use

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

Molecular networking, conformal predictions and revised fingerprint-based models for discovering endocrine disruptors in mixtures.

Analytical and bioanalytical chemistry·2026

Same author

Classification of industrial chemicals for respiratory chemosensory irritation using the TRPV1-expressing neuronal SH-SY5Y cell model and machine learning.

Archives of toxicology·2026

Same author

Prediction of the classification, labelling and packaging regulation H-statements with confidence using conformal prediction with N-grams and molecular fingerprints.

Current research in toxicology·2025

Same author

Rapid traversal of vast chemical space using machine learning-guided docking screens.

Nature computational science·2025

Same author

CPSign: conformal prediction for cheminformatics modeling.

Journal of cheminformatics·2024

Same author

hERG-toxicity prediction using traditional machine learning and advanced deep learning techniques.

Current research in toxicology·2023

Same journal

Correction: Chen et al. Chemical Composition of <i>Litsea pungens</i> Essential Oil and Its Potential Antioxidant and Antimicrobial Activities. <i>Molecules</i> 2023, <i>28</i>, 6835.

Molecules (Basel, Switzerland)·2026

Same journal

Correction: Ruan et al. Comparison of Extraction, Isolation, Purification, Structural Characterization and Immunomodulatory Activity of Polysaccharides from Two Species of <i>Cistanche</i>. <i>Molecules</i> 2025, <i>30</i>, 4754.

Molecules (Basel, Switzerland)·2026

Same journal

Correction: Li et al. Gastrodin Ameliorates Cognitive Dysfunction in Vascular Dementia Rats by Suppressing Ferroptosis via the Regulation of the Nrf2/Keap1-GPx4 Signaling Pathway. <i>Molecules</i> 2022, <i>27</i>, 6311.

Molecules (Basel, Switzerland)·2026

Same journal

Correction: Zueva et al. Steady-State Kinetics of Enzyme-Catalyzed Hydrolysis of Echothiophate, a P-S Bonded Organophosphorus as Monitored by Spectrofluorimetry. <i>Molecules</i> 2020, <i>25</i>, 1371.

Molecules (Basel, Switzerland)·2026

Same journal

1,4-Diazatriphenylene and Its Hetero-Fused Analogs: Synthesis and Applications.

Molecules (Basel, Switzerland)·2026

Same journal

Comparative Phytochemical Studies on the Aerial Parts of <i>Teucrium davaeanum</i> Coss. and <i>Teucrium zanonii</i> Pamp.

Molecules (Basel, Switzerland)·2026

See all related articles

Search research articles

Related Experiment Video

Updated: Aug 15, 2025

A High-throughput Assay for the Prediction of Chemical Toxicity by Automated Phenotypic Profiling of Caenorhabditis elegans

A High-throughput Assay for the Prediction of Chemical Toxicity by Automated Phenotypic Profiling of Caenorhabditis elegans

Published on: March 14, 2019

Traditional Machine and Deep Learning for Predicting Toxicity Endpoints.

¹Department of Computer and Systems Sciences, Stockholm University, 164 07 Kista, Sweden.

Molecules (Basel, Switzerland)

|January 8, 2023

Summary

This summary is machine-generated.

Deep learning models using Simplified Molecular Input Line Entry System (SMILES) show promise for predicting compound toxicity. Bidirectional Encoder Representations from Transformers (BERT) architectures effectively handle imbalanced datasets in drug discovery.

Keywords:

BERT CATMoS dataset CDDD RDKit conformal prediction random forest

More Related Videos

Predicting Treatment Response to Image-Guided Therapies Using Machine Learning: An Example for Trans-Arterial Treatment of Hepatocellular Carcinoma

Predicting Treatment Response to Image-Guided Therapies Using Machine Learning: An Example for Trans-Arterial Treatment of Hepatocellular Carcinoma

Published on: October 10, 2018

Automation of the Micronucleus Assay Using Imaging Flow Cytometry and Artificial Intelligence

Automation of the Micronucleus Assay Using Imaging Flow Cytometry and Artificial Intelligence

Published on: January 27, 2023

Related Experiment Videos

Last Updated: Aug 15, 2025

A High-throughput Assay for the Prediction of Chemical Toxicity by Automated Phenotypic Profiling of Caenorhabditis elegans

A High-throughput Assay for the Prediction of Chemical Toxicity by Automated Phenotypic Profiling of Caenorhabditis elegans

Published on: March 14, 2019

Predicting Treatment Response to Image-Guided Therapies Using Machine Learning: An Example for Trans-Arterial Treatment of Hepatocellular Carcinoma

Predicting Treatment Response to Image-Guided Therapies Using Machine Learning: An Example for Trans-Arterial Treatment of Hepatocellular Carcinoma

Published on: October 10, 2018

Automation of the Micronucleus Assay Using Imaging Flow Cytometry and Artificial Intelligence

Automation of the Micronucleus Assay Using Imaging Flow Cytometry and Artificial Intelligence

Published on: January 27, 2023

Area of Science:

Computational chemistry
cheminformatics
machine learning in drug discovery

Background:

Molecular structure property modeling is crucial for predicting compound characteristics in drug discovery.
Traditional methods are resource-intensive, and toxicity issues cause late-stage attrition.
Deep learning approaches are gaining traction for improved predictive modeling.

Purpose of the Study:

To compare traditional physico-chemical descriptor methods with deep learning approaches for molecular property prediction.
To evaluate descriptor-free, Simplified Molecular Input Line Entry System (SMILES)-based deep learning architectures, specifically Bidirectional Encoder Representations from Transformers (BERT).
To assess the effectiveness of the Mondrian aggregated conformal prediction framework in handling class imbalance.

Main Methods:

Utilized autoencoder-generated descriptors and two descriptor-free, SMILES-based deep learning architectures (BERT type).
Employed the Mondrian aggregated conformal prediction method as an overarching framework.
Tested models on binary CATMoS non-toxic and very-toxic datasets, including an imbalanced dataset with an 11-fold class difference.

Main Results:

All methods performed comparably on the balanced dataset.
The MolBERT model demonstrated superior performance on the imbalanced dataset, achieving high efficiency (0.93-0.94) and balanced accuracy (0.86-0.87).
Descriptor-free BERT architectures produced well-balanced predictive models with defined applicability domains.

Conclusions:

Descriptor-free, SMILES-based BERT architectures are effective for molecular property prediction and toxicity assessment.
The Mondrian conformal prediction method successfully addresses class imbalance without sampling or weighting techniques.
Deep learning, particularly BERT models, offers a powerful, efficient approach to navigating challenges in drug discovery and development.