Scholarlink

<p>This systematic review evaluates the integration of Design of Experiments (DoE) with Artificial Intelligence (AI) and Machine Learning (ML) for developing and validating analytical methods in pharmaceutical and chemical sciences. The literature search (2010–2025) across scientific databases identified peer-reviewed studies combining DoE for data generation with AI/ML for modeling and optimization. The synergy between DoE and ML enhances experimental efficiency, with DoE providing structured data for robust ML models and ML enabling predictive modeling and active learning. Comparative analysis reveals Artificial Neural Networks (ANNs) and Support Vector Machines (SVMs) excel in nonlinear systems, Multiple Linear Regression (MLR) offers interpretable baselines, Adaptive Neuro-Fuzzy Inference Systems (ANFIS) balance accuracy and transparency, Random Forest and XGBoost provide robust ensemble predictions, and Gaussian Process Regression supports uncertainty quantification. Applications include optimizing chromatographic methods, robustness testing, and predictive validation within Analytical Quality by Design (QbD). Challenges involve the “black-box” nature of models, data governance, and Good Manufacturing Practice (GMP) validation. AI (XAI) is emerging to improve transparency. This integration transforms analytical method development, accelerating timelines and supporting autonomous laboratories, with addressing interpretability and regulatory hurdles critical for broader adoption.</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

The convergence of statistical design and computational intelligence: a systematic review of DOE-driven AI/ML in analytical method development and validation

  • 期刊论文

  • 发表日期：

    2026年3月24日

查看全文
  • Huma Sulthana,
  • Judy Jays,
  • Prakash Kumar B

摘要

This systematic review evaluates the integration of Design of Experiments (DoE) with Artificial Intelligence (AI) and Machine Learning (ML) for developing and validating analytical methods in pharmaceutical and chemical sciences. The literature search (2010–2025) across scientific databases identified peer-reviewed studies combining DoE for data generation with AI/ML for modeling and optimization. The synergy between DoE and ML enhances experimental efficiency, with DoE providing structured data for robust ML models and ML enabling predictive modeling and active learning. Comparative analysis reveals Artificial Neural Networks (ANNs) and Support Vector Machines (SVMs) excel in nonlinear systems, Multiple Linear Regression (MLR) offers interpretable baselines, Adaptive Neuro-Fuzzy Inference Systems (ANFIS) balance accuracy and transparency, Random Forest and XGBoost provide robust ensemble predictions, and Gaussian Process Regression supports uncertainty quantification. Applications include optimizing chromatographic methods, robustness testing, and predictive validation within Analytical Quality by Design (QbD). Challenges involve the “black-box” nature of models, data governance, and Good Manufacturing Practice (GMP) validation. AI (XAI) is emerging to improve transparency. This integration transforms analytical method development, accelerating timelines and supporting autonomous laboratories, with addressing interpretability and regulatory hurdles critical for broader adoption.