Lahiq Ahmed A, Alshehri Abdullah A, Alsharif Shaker T
Department of Pharmaceutics, College of Pharmacy, Najran University, Najran, 66262, Saudi Arabia.
Department of Clinical Pharmacy, College of Pharmacy, Taif University, Al Huwaya, Taif, 21944, Saudi Arabia.
Sci Rep. 2025 Jul 17;15(1):26007. doi: 10.1038/s41598-025-11823-z.
The development of continuous pharmaceutical manufacturing is crucial and can be analyzed via advanced computational models. Machine learning is a strong computational paradigm that can be integrated into a continuous process to enhance the drugs' solubility and efficacy. In this research, a simulation method for estimating pharmaceutical solubility was considered in green solvents to develop the idea of continuous pharmaceutical manufacturing. Artificial intelligence strategies were utilized to apply models for fitting several solubility datasets. Using machine learning techniques, the solubility of Clobetasol Propionate (CP) was modeled at temperature values between 308 K and 348 K, and pressures in the range of 12.2 MPa to 35.5 MPa. In this research, two models-a neural network-based model called MLP (Multilayer Perceptron) and a probabilistic model called GPR (Gaussian Process Regression)-along with an ensemble voting model based on these two, were considered for modeling. A GWO (Grey Wolf Optimization) method was also used to tune their hyperparameters. All three models have significant performances on estimation of CP solubility. But the voting model, which is a combination of the other two models, is better than the other two models in terms of accuracy. The ensemble voting model, integrating MLP and GPR with GWO optimization, offers superior accuracy for predicting CP solubility, advancing continuous pharmaceutical manufacturing.
连续制药制造的发展至关重要，并且可以通过先进的计算模型进行分析。机器学习是一种强大的计算范式，可以集成到连续过程中以提高药物的溶解度和疗效。在本研究中，考虑了一种在绿色溶剂中估算药物溶解度的模拟方法，以发展连续制药制造的理念。利用人工智能策略应用模型来拟合多个溶解度数据集。使用机器学习技术，在308 K至348 K的温度值以及12.2 MPa至35.5 MPa的压力范围内对丙酸氯倍他索（CP）的溶解度进行建模。在本研究中，考虑了两种模型——一种基于神经网络的称为MLP（多层感知器）的模型和一种称为GPR（高斯过程回归）的概率模型——以及基于这两种模型的集成投票模型进行建模。还使用了灰狼优化（GWO）方法来调整它们的超参数。所有三种模型在CP溶解度估计方面都有显著表现。但作为其他两种模型组合的投票模型在准确性方面优于其他两种模型。将MLP和GPR与GWO优化相结合的集成投票模型在预测CP溶解度方面具有更高的准确性，推动了连续制药制造的发展。
