Predicting the Efficiency of Nanoliposomes in Cancer Treatment Using Machine Learning Approach Publisher



Ahmadi M ; Davarikia K ; Azari S ; Pouyanfar N ; Masoumi N ; Ghorbani Bidkorpeh F ; Alizadeh B ; Ayyoubzadeh S M
Source: Journal of Pharmaceutical Innovation Published:2027

Abstract

Purpose: Nanoliposomes offer tremendous advantages in precise cancer therapy. However, complex design of liposomal formulations, depending on many physicochemical and biological parameters, renders empirical optimization time-consuming. Machine learning (ML) presents an efficient solution to manage such complexity and optimize drug delivery systems. Methods: In the current study, an optimal ML workflow was constructed utilizing a dataset containing 786 records from 100 research articles. In this study, five regression algorithms were selected, including Gradient Boosting, Linear Regression, Random Forest, Support Vector Regression (SVR), and XGBoost. The performance of the models was carefully evaluated using 5-fold cross-validation based on Mean Squared Error (MSE), Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), R-squared (R2), and adjusted R-squared (Adj. R2) metrics. Furthermore, SHapley Additive exPlanations (SHAP) was employed for model interpretability. Results: Results suggested that ensemble models may capture patterns in the literature-derived dataset; however, given the limited effective sample sizes after preprocessing, these findings should be interpreted as exploratory and potentially optimistic. Gradient Boosting provided an R2 of 0.905 (Adj. R2 = 0.855), MSE of 50.70, MAE of 5.97, and RMSE of 7.12 for cell viability and Random Forest provided an R2 of 0.963 (Adj. R2 = 0.883), MSE of 0.0007, MAE of 0.0120, and RMSE of 0.0266 for IC50 (half-maximal inhibitory concentration). SVR demonstrated the best performance (R2 of 0.614, Adj. R2 = 0.158, MSE of 35778, MAE of 93.4, and RMSE of 189.1) for in vivo tumor growth. Furthermore, Linear Regression exhibited regular low performance consistently. Furthermore, SHAP analysis exhibited major effective features such as concentration, size, zeta potential, and loading efficiency on target variables, offering significant information regarding an optimal nanoliposome design. Conclusion: This ML-based approach provides a preliminary and exploratory framework for organizing nanoliposomal formulation data and identifying candidate variables that may influence therapeutic efficacy. Larger datasets and external validation are required before reliable formulation optimization can be claimed. shifting from empirical methods to accelerate the development of highly efficient and customized cancer therapy. © The Author(s), under exclusive licence to Springer Science+Business Media, LLC, part of Springer Nature 2026.