Scholarlink

Objective <p>To evaluate artificial neural networks (ANNs) as an AI-driven tool for predicting and optimizing the dissolution performance of etodolac extended-release (ER) tablets.</p> Methods <p>A 3-factor, 3-level full factorial design (27 runs) was prepared by varying Hydroxypropyl Methylcellulose (HPMC K100M; 120–170&#xa0;mg/tablet), Ethyl Cellulose (Premium CR 7 cps; 15–55&#xa0;mg/tablet), and Dibasic Calcium Phosphate (DCP; 50–90&#xa0;mg/tablet) using a high-shear wet granulation process. Dissolution was measured using USP Apparatus II (75&#xa0;rpm) in pH 6.8 phosphate buffer, and multipoint profiles were generated (primary discriminatory time points: 2, 4, and 8&#xa0;h). Data were randomly split into training/validation/test subsets (70/10/20) to screen multiple machine-learning models and train a boosted ANN.</p> Results <p>The ANN provided the most reliable predictive framework among the screened methods and showed strong goodness-of-fit for early and mid-time points (2, 4 and 8&#xa0;h) with consistent performance across holdout datasets. Variable-importance and profiler analyses indicated ethyl cellulose and HPMC K100M as the dominant factors impacting drug release, with DCP showing a comparatively smaller effect. A desirability-based optimization and design-space/contour profiling identified formulation regions meeting USP Test 3 dissolution targets.</p> Conclusion <p>ANN-driven modeling can capture nonlinear excipient–performance relationships and enables efficient, data-driven optimization of etodolac ER tablets, reducing experimental burden while supporting QbD-aligned formulation development.</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

Advancing Formulation Development with AI: ANN-Driven Modeling and Optimization of Etodolac Extended-Release Tablets

  • 期刊论文

  • 发表日期：

    2026年3月18日

查看全文
  • Paritosh Kunwar Singh,
  • Vinay Kumar,
  • Ashish Parihar,
  • Gauri Goyal,
  • Suruchi Singh,
  • Pankaj Bhatt

摘要

Objective

To evaluate artificial neural networks (ANNs) as an AI-driven tool for predicting and optimizing the dissolution performance of etodolac extended-release (ER) tablets.

Methods

A 3-factor, 3-level full factorial design (27 runs) was prepared by varying Hydroxypropyl Methylcellulose (HPMC K100M; 120–170 mg/tablet), Ethyl Cellulose (Premium CR 7 cps; 15–55 mg/tablet), and Dibasic Calcium Phosphate (DCP; 50–90 mg/tablet) using a high-shear wet granulation process. Dissolution was measured using USP Apparatus II (75 rpm) in pH 6.8 phosphate buffer, and multipoint profiles were generated (primary discriminatory time points: 2, 4, and 8 h). Data were randomly split into training/validation/test subsets (70/10/20) to screen multiple machine-learning models and train a boosted ANN.

Results

The ANN provided the most reliable predictive framework among the screened methods and showed strong goodness-of-fit for early and mid-time points (2, 4 and 8 h) with consistent performance across holdout datasets. Variable-importance and profiler analyses indicated ethyl cellulose and HPMC K100M as the dominant factors impacting drug release, with DCP showing a comparatively smaller effect. A desirability-based optimization and design-space/contour profiling identified formulation regions meeting USP Test 3 dissolution targets.

Conclusion

ANN-driven modeling can capture nonlinear excipient–performance relationships and enables efficient, data-driven optimization of etodolac ER tablets, reducing experimental burden while supporting QbD-aligned formulation development.

Graphical Abstract