Preparation of solid-dosage nanomedicine via green chemistry route: Advanced computational simulation of nanodrug solubility prediction using machine learning models

We have developed a novel methodology for estimating the solubility of Letrozole (LET) in a green solvent for development of green-based nanomedicine manufacture. The method employed in this research does not utilize chemical solvents for the processing of drug particles, thereby improving the process efficiency. The solubility of LET drug has been investigated based on two parameters, temperature and pressure due to their significant influence on the drug solubility behavior. Classification and regression trees (CART), Gaussian process regression (GPR), and multilayer perceptron (MLP) are selected models for this investigation to determine the solubility values. Also, Jellyfish optimizer (JO) is used for model optimization and hyper-parameters tuning. After obtaining the tuned models, they were evaluated using different statistical metrics and the results were compared. CART, GPR, and MLP had R2 scores of 0.9192, 0.9681, and 0.9736 in test phase, respectively. The MLP model was selected as the most accurate and general model in estimating Letrozole in the solvent. The optimized model had error rates of 0.1133, 0.1503, and 0.25 in terms of MAE, RMSE, and Max Error, respectively.