Our study delves into the exciting realm of transdermal drug delivery, a growing preference for patients due to its convenience. Recognizing the challenge posed by the skin's natural barrier to drug permeation, we employed advanced machine learning models to predict skin permeability solely based on descriptors computationally calculated from the chemical structure of the molecule.
This repository contains code and resources for predicting skin permeability of drug molecules using machine learning models. The predictive models are trained on a diverse dataset of molecules, encompassing drugs, xenobiotics, and other chemical compounds.
- Predictive Models: Utilize advanced AI algorithms, including MLR, ensemble methods, and artificial neural networks.
- Dataset: In vitro human skin permeation parameters for diverse molecules, including drugs and xenobiotics.
- Molecular Descriptors: Calculated using the Chemistry Development Kit (CDK) for one-dimensional and two-dimensional representations.
Clone the repository to your local machine using:
conda create --name myenv python=3.10
conda activate myenv
pip install -r requirements.txt
The dataset used in this study was obtained from Cheruvu et al., providing in vitro human skin permeation parameters for a diverse range of molecules.
- Molecular Descriptor Calculation: Employ the CDK to generate descriptors for molecules.
- AI Models: Develop regression models using Scikit-Learn with algorithms such as MLR, ensemble methods, and artificial neural networks.
- Cluster Analysis: Predict FDA-approved drug permeability, conduct K-means clustering, and use the Anatomical Therapeutic Chemical (ATC) code for drug classification.
Contributions are welcome! Fork the repository, make your changes, and submit a pull request.
The study acknowledges the work of Cheruvu et al. for providing the foundational skin permeability dataset. We acknowledge DrugBank for providing the dataset of FDA-approved drugs.
This project is licensed under the MIT License.