Prediction and modeling of physicochemical properties of surfactants of pharmaceutical interest

Abstract

Amphiphiles, also called surfactants, are widely known for their ability to reduce surface tension and are commonly used in different industrial fields due to their interesting physicochemical characteristics. This work aimed to develop and evaluate QSPR models that contribute to the modeling of surfactant properties using machine learning algorithms. The first contribution focused on predicting the critical micelle concentration (CMC) of 119 sugar-based surfactants by applying four statistical learning techniques including Multiple linear regression, Partial least square, Artificial neural networks (ANN), and Adaptive neuro-fuzzy inference system. Different statistical metrics were employed to evaluate the reliability and robustness of the models. The best result (𝑟̅ 𝑚2= 0.803, 𝑄𝐹12=0.982, and Δ𝑟𝑚2= 0.006) were obtained for ANN with {6-6-1} architecture. In addition, 6 other sugar-based surfactants were selected as external sets to assess the developed ANN model and the result gave good prediction with (R = 0.96). The second contribution focused on predicting the critical micelle concentration (pCMC) of different classes of surfactants, including anionic, cationic, nonionic, zwitterionic, and Gemini surfactants. Statistical analysis revealed that a set of 14 significant Mordred descriptors (SlogP, GATS6d, nAcid, GATS8dv, GATS4dv, PEOE_VSA11, GATS8d, ATS0p, GATS1d, MATS5p, GATS3d, NdssC, GATS6dv and EState_VSA4), along with temperature, served as appropriate inputs. Different machine learning methods, such as MLR, random forest regression (RFR), ANN, and optimized Support Vector Regression by the dragonfly algorithm (DA-SVR) model their predictive/correlative capacity and feasibility have been examined. According to the statistical coefficients, DA-SVR was the most accurate in predicting pCMC values, achieving (𝑅2=0.9740, 𝑄2=0.9739, 𝑟̅ 𝑚2=0.9627, and Δ𝑟𝑚2=0.0244) for the entire dataset