Analysis of nanomedicine production via green processing: Modeling and simulation of pharmaceutical solubility using artificial intelligence

Abstract

This research focuses on investigating the solubility of tolfenamic acid in SC-CO2 (supercritical carbon dioxide) and the density of SC-CO2 solvent via theoretical artificial intelligence method. The study involves analyzing the relationship between temperature, pressure, and the corresponding mentioned outputs for the process. Three different predictive models, namely Multi-layer Perceptron (MLP), Polynomial Regression (PR), and Extra Trees (ET) are utilized to forecast the solubility of drug and the density of solvent (SC–CO2). The models are fine-tuned with hyper-parameters using the Dragonfly Algorithm (DA) to ensure accurate predictions. The solubility prediction is remarkably accurate using the MLP model, showing a high score of 0.98329 in terms of R-squared. The maximum error is 0.2474, and the MAE is 0.1095, demonstrating the model's high precision in estimating tolfenamic acid's solubility in SC-CO2. The PR model demonstrates exceptional accuracy, yielding a score of 0.99844 by R-squared metric, a maximum error of 0.068, and an MAE of 0.0314. The ET model also performs well, with an R-squared score of 0.90977, a maximum error of 0.445, and an MAE of 0.1665. Regarding density prediction, MLP outperforms the other techniques, achieving a significant R2 parameter of 0.99919, an MSE of 12.663, and a mean MAPE of 0.0037.