Related Experiment Video
Updated: Aug 5, 2026
11:51
Systemic and Local Drug Delivery for Treating Diseases of the Central Nervous System in Rodent Models
Published on: August 16, 2010
Comparative analysis of supervised machine learning algorithms for transdermal drug delivery in brain disorders
Hetalbahen Kiritkumar Dave1, Tejas Harshadbhai Thakkar2, Vaishali Tejas Thakkar3
1Department of Computer Science, The Charutar Vidya Mandal (CVM) University, Vallabh Vidyanagar, Gujarat, 388120, India.
Journal of Computer-Aided Molecular Design
|August 4, 2026
Summary
Machine learning models predict key attributes for brain-targeted transdermal drug delivery systems (TDDS), accelerating development. Ensemble models significantly improved prediction accuracy for formulation efficiency and drug release.
Area of Science:
- Pharmacology and Pharmaceutics
- Biomedical Engineering
- Computational Science
Background:
- Transdermal drug delivery systems (TDDS) offer non-invasive treatment for brain diseases, improving patient compliance and drug stability.
- Optimizing TDDS is challenging due to complex interactions between formulation components and process variables affecting drug delivery.
- Current methods lack efficiency in predicting multiple critical formulation attributes simultaneously for brain-targeted TDDS.
Purpose of the Study:
- To develop a supervised Machine Learning (ML) framework for simultaneously predicting multiple critical formulation attributes in brain-targeted TDDS.
- To create an accurate, interpretable, and scalable decision-support platform for accelerating TDDS development.
Main Methods:
- Utilized a dataset of 542 formulation records from 48 peer-reviewed studies and 6 laboratory sources.
- Integrated physiochemically informed preprocessing, feature engineering, and Cuckoo Catfish Optimizer (CCO)-based hyperparameter tuning with various supervised learning models.
- Employed SHapley Additive exPlanations (SHAP) for simultaneous prediction and interpretation of seven key formulation outputs.
Main Results:
- Ensemble ML models significantly outperformed linear models, showing an average R increase of 24.63% in prediction accuracy.
- SHAP analysis identified lipid composition, surfactant type, pH, and preparation temperature as key determinants for successful TDDS formulations.
- The developed framework achieved comparable stability and performance across most formulation properties.
Conclusions:
- The supervised ML framework provides an accurate, interpretable, and scalable decision-support tool for TDDS development.
- This data-driven approach significantly reduces experimental workload and accelerates the design of effective brain-targeted TDDS.
- The findings highlight the potential of ML in optimizing complex pharmaceutical formulations for neurological treatments.