Scholarlink

Purpose <p>Pharmaceutical nanoparticle involves complex and highly nonlinear interactions between formulation variables and process parameters that are difficult to describe using conventional statistical models. Artificial intelligence (AI) offers an advanced approach to overcoming these limitations and enhancing formulation optimization and quality control. This study aimed to develop an AI-driven framework integrating hybrid machine learning (HML) and a reinforcement learning–enhanced genetic algorithm (GA-RL) to optimize resveratrol-loaded polymeric nanoparticles (RES-PNPs) for cancer treatment.</p> Methods <p>A dataset of RES-PNPs was used to train ML, including linear regression, k-nearest neighbors, support vector machines, and artificial neural networks. HML was constructed to predict particle size (PS), polydispersity index (PDI), zeta potential (ZP), and percentage encapsulation efficiency (%EE). An AI-based multidimensional design space was established, followed by optimization using GA-RL.</p> Results <p>The optimized formulation was validated and evaluated physicochemically and biologically. The results reported that hybrid ML models achieved high predictive accuracy. GA-RL optimization identified an optimal formulation with experimentally validated values of PS 79.58 ± 8.53 nm, PDI 0.40 ± 0.05, ZP −49.60 ± 1.25 mV, and %EE 70.65 ± 1.52%, with no significant differences from the predicted values. Optimized RES-PNPs in 5%trehalose exhibited sustained and pH-responsive drug release, enhanced cellular uptake, increased ROS generation, and apoptosis-mediated cytotoxicity against skin cancer cells, while maintaining biocompatibility with normal fibroblasts.</p> Conclusion <p>This study demonstrated novel AI-based HML–GA-RL framework capable of accurately modeling complex formulation relationships and enabling intelligent construction of the design space, offering a powerful strategy for advanced pharmaceutical formulation and nanomedicine quality control.</p> Graphical abstract <p></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

Development of hybrid machine learning models integrated with reinforcement learning–enhanced genetic algorithms for quality control and optimization of resveratrol-loaded polymeric nanoparticles in cancer treatment

  • 期刊论文

  • 发表日期：

    2026年5月8日

查看全文
  • Phuvamin Suriyaamporn,
  • Boonnada Pamornpathomkul,
  • Tanasait Ngawhirunpat,
  • Praneet Opanasopit,
  • Warut Pannakkong

摘要

Purpose

Pharmaceutical nanoparticle involves complex and highly nonlinear interactions between formulation variables and process parameters that are difficult to describe using conventional statistical models. Artificial intelligence (AI) offers an advanced approach to overcoming these limitations and enhancing formulation optimization and quality control. This study aimed to develop an AI-driven framework integrating hybrid machine learning (HML) and a reinforcement learning–enhanced genetic algorithm (GA-RL) to optimize resveratrol-loaded polymeric nanoparticles (RES-PNPs) for cancer treatment.

Methods

A dataset of RES-PNPs was used to train ML, including linear regression, k-nearest neighbors, support vector machines, and artificial neural networks. HML was constructed to predict particle size (PS), polydispersity index (PDI), zeta potential (ZP), and percentage encapsulation efficiency (%EE). An AI-based multidimensional design space was established, followed by optimization using GA-RL.

Results

The optimized formulation was validated and evaluated physicochemically and biologically. The results reported that hybrid ML models achieved high predictive accuracy. GA-RL optimization identified an optimal formulation with experimentally validated values of PS 79.58 ± 8.53 nm, PDI 0.40 ± 0.05, ZP −49.60 ± 1.25 mV, and %EE 70.65 ± 1.52%, with no significant differences from the predicted values. Optimized RES-PNPs in 5%trehalose exhibited sustained and pH-responsive drug release, enhanced cellular uptake, increased ROS generation, and apoptosis-mediated cytotoxicity against skin cancer cells, while maintaining biocompatibility with normal fibroblasts.

Conclusion

This study demonstrated novel AI-based HML–GA-RL framework capable of accurately modeling complex formulation relationships and enabling intelligent construction of the design space, offering a powerful strategy for advanced pharmaceutical formulation and nanomedicine quality control.

Graphical abstract