SimSJSAlert: A Similarity-Augmented Multi-View Learning Framework with Scaffold Alerts for Drug-Induced Stevens-Johnson Syndrome Risk Assessment
5 affiliations
- 1Faculty of Health Sciences, College of Natural Sciences
- 2Can Tho University
- 3Graduate School in the Program of Research and Development in Pharmaceuticals, Faculty of Pharmaceutical Sciences
- 4Khon Kaen University
- 5Division of Pharmaceutical Chemistry, Faculty of Pharmaceutical Sciences
Abstract
Stevens-Johnson syndrome (SJS) is a rare, life-threatening adverse drug reaction for which early identification of high-risk drugs remains challenging. In this study, we developed an explainable quantitative structure-toxicity relationship (QSTR) framework for SJS prediction using a multi-view learning framework with 21 molecular representations, including molecular descriptors, fingerprints, transformer-based embeddings, and a molecular similarity representation. Model development and evaluation were designed in accordance with the OECD principles for QSAR validation, incorporating scaffold-aware validation, external evaluation on pharmacovigilance negative-signal compounds, and structural interpretation by integrating SHAP with Bemis-Murcko scaffold enrichment analysis. The proposed framework achieved robust sensitivity and specificity, with molecular similarity consistently improving the identification of SJS-associated compounds. Furthermore, a consensus prediction and scaffold-based prioritization strategy identified several high-risk drugs from the 1 external negative-signal dataset, including cefprozil, tetracycline, and omadacycline, highlighting their potential association with SJS and warranting further pharmacovigilance and experimental investigation. To facilitate practical application, this was deployed as a freely accessible web server, SimSJSAlert (https: //sim-sjs-alert.streamlit.app/), enabling rapid drug-induced SJS risk assessment. The proposed framework combines predictive performance, interpretability, and accessibility to support early SJS risk assessment, pharmacovigilance signal prioritization, and safer drug development.
Related
- Drug Molecular and Route Similarity-Aware Retrieval-Augmented LLM Generation with Knowledge Graph for Semantic Reasoning
- An Expert-Augmented Deep Learning Approach for Synthesis Route Evaluation
- A predictive transferability index for crystal plasticity framework selection: Variational derivation, analytical bounds, and FFT-based plausibility assessment
- Application of spectroscopic and multispectral imaging technologies on the assessment of ready-to-eat pineapple quality: A performance evaluation study of machine learning models generated from two commercial data analytics tools
- A Dynamic Residual Value Assessment and Trade-In Decision Framework for Retired Power Batteries Targeting Multi-Scenario Secondary Utilization
BibTeX
@misc{duy2026simsjsalert,
title = {{SimSJSAlert: A Similarity-Augmented Multi-View Learning Framework with Scaffold Alerts for Drug-Induced Stevens-Johnson Syndrome Risk Assessment}},
author = {Huynh Anh Duy and Sastiya Kampaengsri and Supreeya Paiboon and Tarapong Srisongkram},
howpublished = {ChemRxiv},
year = {2026},
month = sep,
doi = {10.26434/chemrxiv.15008910/v1},
url = {https://chemrxiv.org/doi/full/10.26434/chemrxiv.15008910/v1},
note = {CC BY-NC},
}
RIS
TY - UNPB
TI - SimSJSAlert: A Similarity-Augmented Multi-View Learning Framework with Scaffold Alerts for Drug-Induced Stevens-Johnson Syndrome Risk Assessment
AU - Huynh Anh Duy
AU - Sastiya Kampaengsri
AU - Supreeya Paiboon
AU - Tarapong Srisongkram
PY - 2026
DA - 2026/09/16
PB - ChemRxiv
DO - 10.26434/chemrxiv.15008910/v1
UR - https://chemrxiv.org/doi/full/10.26434/chemrxiv.15008910/v1
AB - Stevens-Johnson syndrome (SJS) is a rare, life-threatening adverse drug reaction for which early identification of high-risk drugs remains challenging. In this study, we developed an explainable quantitative structure-toxicity relationship (QSTR) framework for SJS prediction using a multi-view learning framework with 21 molecular representations, including molecular descriptors, fingerprints, transformer-based embeddings, and a molecular similarity representation. Model development and evaluation were designed in accordance with the OECD principles for QSAR validation, incorporating scaffold-aware validation, external evaluation on pharmacovigilance negative-signal compounds, and structural interpretation by integrating SHAP with Bemis-Murcko scaffold enrichment analysis. The proposed framework achieved robust sensitivity and specificity, with molecular similarity consistently improving the identification of SJS-associated compounds. Furthermore, a consensus prediction and scaffold-based prioritization strategy identified several high-risk drugs from the 1 external negative-signal dataset, including cefprozil, tetracycline, and omadacycline, highlighting their potential association with SJS and warranting further pharmacovigilance and experimental investigation. To facilitate practical application, this was deployed as a freely accessible web server, SimSJSAlert (https: //sim-sjs-alert.streamlit.app/), enabling rapid drug-induced SJS risk assessment. The proposed framework combines predictive performance, interpretability, and accessibility to support early SJS risk assessment, pharmacovigilance signal prioritization, and safer drug development.
ER -