Scholarlink

<p>Lacosamide, a widely used antiepileptic drug, suffers from poor solubility in conventional solvents, which limits its bioavailability. Supercritical carbon dioxide (SC-CO₂) has emerged as an environmentally friendly substitute solvent for pharmaceutical processing. In this study, the solubility of Lacosamide in SC-CO₂ was modeled and predicted using several machine learning techniques, including Gradient Boosting Decision Tree (GBDT), Multilayer Perceptron (MLP), Random Forest (RF), Gaussian Process Regression (GPR), Extreme Gradient Boosting (XG Boost), and Polynomial Regression (PR). These models have the ability to model nonlinear relationships. Experimental solubility information within a large span of pressures and temperatures were employed for model training and validation. The findings suggested that all applied models were competent in providing reliable predictions, with GBDT (R<sup>2</sup> = 0.9989), XG Boost (R<sup>2</sup> = 0.9986), and MLP (R<sup>2</sup> = 0.9975) exhibiting the highest accuracy, achieving the highest coefficient of determination (R<sup>2</sup>). Overall, combining experimental data with advanced machine learning algorithms offers a powerful approach for predicting and optimizing drug solubility in supercritical systems, thereby facilitating the design of scalable pharmaceutical processes.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Investigation of Lacosamide solubility in supercritical carbon dioxide with machine learning models

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年11月27日

查看全文
  • Nadia Esfandiari

摘要

Lacosamide, a widely used antiepileptic drug, suffers from poor solubility in conventional solvents, which limits its bioavailability. Supercritical carbon dioxide (SC-CO₂) has emerged as an environmentally friendly substitute solvent for pharmaceutical processing. In this study, the solubility of Lacosamide in SC-CO₂ was modeled and predicted using several machine learning techniques, including Gradient Boosting Decision Tree (GBDT), Multilayer Perceptron (MLP), Random Forest (RF), Gaussian Process Regression (GPR), Extreme Gradient Boosting (XG Boost), and Polynomial Regression (PR). These models have the ability to model nonlinear relationships. Experimental solubility information within a large span of pressures and temperatures were employed for model training and validation. The findings suggested that all applied models were competent in providing reliable predictions, with GBDT (R2 = 0.9989), XG Boost (R2 = 0.9986), and MLP (R2 = 0.9975) exhibiting the highest accuracy, achieving the highest coefficient of determination (R2). Overall, combining experimental data with advanced machine learning algorithms offers a powerful approach for predicting and optimizing drug solubility in supercritical systems, thereby facilitating the design of scalable pharmaceutical processes.