%0 Journal Article
%T 基于机器学习的hERG心脏毒性预测模型：利用分子指纹特征提高药物安全性评估
The hERG Cardiotoxicity Prediction Model Based on the Machine Learning: Enhancing Drug Safety Assessment by Using Molecular Fingerprints Features
%A 沈哲兴
%A 朱子墨
%A 王立佐
%A 蔡慧芝
%A 胡雅静
%A 叶胜星
%A 焦佳丽
%J Bioprocess
%P 22-28
%@ 2164-5582
%D 2025
%I Hans Publishing
%R 10.12677/bp.2025.151004
%X 本研究旨在探讨基于机器学习的心脏毒性预测模型，特别是针对hERG心脏毒性的预测。hERG心脏毒性是药物开发中的关键安全问题，可能导致QT间期延长综合征，增加心律失常的风险。通过机器学习方法，利用分子指纹特征对化合物的心脏毒性进行预测。本研究的主要发现包括：通过逻辑回归、随机森林、支持向量机和神经网络等算法建立的预测模型，能够准确预测hERG心脏毒性，为药物安全性评估提供了一种新的数据驱动方法。结果表明，随机森林模型的预测性能最佳，准确率达到85%，显示出其在药物安全性评估中的应用潜力。此外，SVM和MLP模型的准确率也较高，而逻辑回归模型的泛化能力相对较弱。本研究为心脏毒性预测提供了一种数据驱动的方法，有助于提高药物开发的安全性和效率。
This study aims to explore machine learning-based cardiotoxicity prediction models, particularly for predicting hERG cardiotoxicity. hERG cardiotoxicity is a critical safety issue in drug development, as it can lead to QT interval prolongation syndrome and increase the risk of arrhythmia. Using machine learning methods, we predict the cardiotoxicity of compounds based on molecular fingerprint features. The key findings of this study include the development of predictive models using logistic regression, random forest, support vector machines, and neural networks, which accurately predict hERG cardiotoxicity. This provides a novel data-driven approach for drug safety assessment. Results show that the RF model achieves the best predictive performance with an accuracy of 85%, demonstrating its potential application in drug safety assessment. Additionally, the SVM and MLP models also exhibit high accuracy, while the LR model has relatively poor generalization ability. This study provides a data-driven method for predicting cardiotoxicity, contributing to improving drug development safety and efficiency.
%K 机器学习，
%K 分子指纹特征，
%K 逻辑回归
Machine Learning
%K Molecular Fingerprinting Features
%K Logistic Regression
%U http://www.hanspub.org/journal/PaperInformation.aspx?PaperID=108921