Cardiosim-Tox: an interpretable multitask deep learning QSAR platform with multimodal feature fusion for predicting hERG, Cav1.2, and Nav1.5 blockade risk and potency
Archives of Toxicology · 2026-12 · 100 (12) · 5571-5609
DOI: 10.1007/s00204-026-04519-4 ↗Kurz gefasst
Fachartikel in der Zeitschrift Archives of Toxicology, online erschienen am 11.08.2026. Verfasst von Fauzan Syarif Nursyafi und 9 weiteren Autor:innen. Beteiligte Einrichtung: Department of Biomedical Engineering, Kumoh National Institute of Technology, Gumi, 39177, Republic of Korea. Hauptthemen laut MEDLINE: Quantitative Structure-Activity Relationship, ERG1 Potassium Channel, Deep Learning, Calcium Channel Blockers. Die Arbeit nennt 78 Literaturangaben. Gefördert von: National Research Foundation of Korea under the Basic Science Research Program, Ministry of Science and the Institute of Information & Communications Technology Planning & Evaluation (IITP) through the Innovative Human Resource Development…, Korea Innovation Foundation (INNOPOLIS) grant funded by the Korea government (MSIT).
Abstract (Originalsprache)
Drug-induced cardiotoxicity, mainly driven by cardiac ion-channel blockade, remains a leading cause of drug attrition and post-market withdrawal, highlighting the need for reliable early-stage screening tools. Existing computational methods, including QSAR models, largely focus on single ion channels, limiting their ability to assess multi-channel safety profiles. To address this gap, we developed Cardiosim-Tox, a modular multi-modal deep learning platform to simultaneously predicts blockade risk (binary classification) and potency (pIC50) for hERG, Cav1.2, and Nav1.5. The framework integrates topological fingerprints, Mordred 2D/3D descriptors, and molecular graph encodings under both single-task (STL) and multi-task learning (MTL) configurations. Developed on 31,816 curated unique compounds, model performance was evaluated on three Tanimoto similarity-stratified external validation sets, with domain-shift metrics and applicability domain analyses supporting robustness. MTL models consistently outperformed STL, while the full modality combination (FP + MD + Graph) achieved AUCs of 0.93-0.98 for Cav1.2, 0.88-0.96 for Nav1.5, and 0.79-0.95 for hERG across validation sets. Regression tasks similarly favored MTL (R2 = 0.94), with multimodal configurations outperforming single-modality baselines. Further comparisons with classical machine learning models and previously developed cardiotoxicity models demonstrated the superiority of Cardiosim-Tox across all endpoints. SHAP analysis confirmed that feature-activity relationships, both at global and mechanistic levels, align with known ion-channel pharmacophores, enhancing interpretability beyond predictive accuracy. Overall, Cardiosim-Tox provides a reliable and interpretable platform for integrated multi-channel cardiac safety assessment, supporting early-stage cardiotoxicity screening in drug development in line with the CiPA paradigm. The Cardiosim-Tox web server is accessible at https://metaheart.kr/login .
Schlagwörter
MeSH-Schlagwörter (MEDLINE)
Stichwörter der Autor:innen
Substanzen
Autor:innen und Institutionen
- Department of Biomedical Engineering, Kumoh National Institute of Technology, Gumi, 39177, Republic of Korea.
- Bioinformatics Research Center, Institute of Bioinformatics Indonesia, Malang, 65162, Indonesia.
- Department of IT Convergence Engineering, Kumoh National Institute of Technology, Gumi, 39177, Republic of Korea.
- Meta Heart Co., Ltd, Gumi, 39253, Republic of Korea.
- Department of Biomedical Engineering, Kumoh National Institute of Technology, Gumi, 39177, Republic of Korea. kmlim@kumoh.ac.kr.
- Meta Heart Co., Ltd, Gumi, 39253, Republic of Korea. kmlim@kumoh.ac.kr.
Förderung
- National Research Foundation of Korea under the Basic Science Research Program
- Ministry of Science and the Institute of Information & Communications Technology Planning & Evaluation (IITP) through the Innovative Human Resource Development…
- Korea Innovation Foundation (INNOPOLIS) grant funded by the Korea government (MSIT)
- Regional Innovation System & Education (RISE)-Regional Growth Innovation LAB program through the Gyeongbuk RISE Center, funded by the Ministry of Education (MO…
Zitieren
APA
Nursyafi, F. S., Kang, B., Eom, J., Izza, R. N., Fauzan, A. L., Putri, N. H., Hanum, U. L., Olivia, Y., Rose, S. A., & Lim, K. M. (2026). Cardiosim-Tox: an interpretable multitask deep learning QSAR platform with multimodal feature fusion for predicting hERG, Cav1.2, and Nav1.5 blockade risk and potency. Archives of Toxicology, 100(12), 5571-5609. https://doi.org/10.1007/s00204-026-04519-4
Vancouver
Nursyafi FS, Kang B, Eom J, Izza RN, Fauzan AL, Putri NH, et al. Cardiosim-Tox: an interpretable multitask deep learning QSAR platform with multimodal feature fusion for predicting hERG, Cav1.2, and Nav1.5 blockade risk and potency. Archives of Toxicology. 2026;100(12):5571-5609. doi:10.1007/s00204-026-04519-4
BibTeX
@article{Nursyafi2026,
author = {Nursyafi, Fauzan Syarif and Kang, Byunggyu and Eom, Junhyeok and Izza, Rahmafatin Nurul and Fauzan, Abdul Latif and Putri, Nadilla Hafani and Hanum, Ulfa Latifa and Olivia, Yossi and Rose, Sanjida Afrin and Lim, Ki Moo},
title = {Cardiosim-Tox: an interpretable multitask deep learning QSAR platform with multimodal feature fusion for predicting hERG, Cav1.2, and Nav1.5 blockade risk and potency},
journal = {Archives of Toxicology},
year = {2026},
volume = {100},
number = {12},
pages = {5571--5609},
doi = {10.1007/s00204-026-04519-4}
}
EUVIMED