Predicting cellular uptake of metal-organic frameworks using machine learning tools

Background and Aims: Metal-organic frameworks have been considered promising candidates for biomedical uses due to their unique characteristics. Predicting cellular uptake of metal-organic frameworks is necessary in order to enhance their design and efficacy. In this study, machine learning models were developed to predict the cellular uptake of metal-organic frameworks and identify significant features influencing uptake efficiency. Methods: In this study, we collected a dataset of studies related to cellular uptakes of various metal-organic frameworks, consisting of ۹۸ records. We used two Machine Learning models, Random Forest and Artificial Neural Networks, to predict the cellular uptake. Feature importance was approximated by correlation analysis, with the target variable being cellular uptake. The default hyperparameters were utilized for the Random Forest model, and the Artificial Neural Networks architecture had three hidden layers (۱۰, ۲۰, and ۱۰ neurons, respectively). Model performance was assessed using Root Mean Squared Error and Mean Absolute Error. We used Leave-One-Out Cross Validation to assess the performance of the models. Results: Correlation analysis showed that the significant features were targeting ligand (۰.۷۴۹), zeta potential (۰.۶۵۸), and coating (۰.۶۵۵), while incubation time (۰.۰۷۶) had an insignificant impact. The Root Mean Squared Error and Mean Absolute Error of the Random Forest model were ۱۵.۸۰۰. The Artificial Neural Networks outperformed Random Forest with a Root Mean Squared Error of ۱۴.۸۰۲ and a Mean Absolute Error of ۱۴.۸۰۲. Conclusion: This study successfully developed Machine Learning models to predict cellular uptake of metal-organic frameworks, where significant importance has been assigned to targeting ligand, zeta potential, and coating as predictors. The Artificial Neural Networks model worked best, suggesting its potential in the optimization of metal-organic framework design for biomedical applications. These findings will be helpful in directing researchers toward the betterment of metal-organic framework-based drug