A novel experimental method for adsorption of fatty acids from pumpkin seed oil in the presence of iron oxide nanoparticles: Experimental and SA – LOOCV – GRBF mathematical modeling

Highlights

• Safe and controllable process was employed to coat fatty acids on the Fe₃O₄ NP.

• Saponification reaction using solvent with natural source was used.

• New hybrid algorithm called SA-LOOCV-GRBF was used to optimize process conditions.

• Average diameter of NPs was 20 nm with a thickness of 5 nm for fatty acid coating.

Abstract

Fatty acids (FA) of pumpkin seed oil have a wide range of applications in treating prostate and skin cancers. The loading of these materials on magnetic nanoparticles (MNPs) constitutes a new approach to drug delivery systems. In this study, a safe and controllable process for adsorption of FAs from pumpkin seed oil on iron oxide nanoparticles has been presented. To extract and adsorb FAs from seed oil, the saponification reaction of oil was carried out along with the innovated hydrothermal method using the evaporation technique in the presence of MNPs. The impact of various parameters such as temperature, distilled water volume, ratio of FA to iron moles, the effect of adsorption process time on FA adsorption value and MNP size were investigated experimentally. Afterward, the optimum conditions to achieve high adsorption efficiency and low particle size were investigated using a novel hybrid SA-LOOCV-GRBF neural network modeling. This method estimated parameters of generalized radial basis function (GRBF) with simulated annealing (SA) optimization and leave one out cross validation (LOOCV) technique. Using LOOCV criteria and SA, the algorithm generated a structure distinct from other methods and achieved optimum parameters without gradient method. It provided a function that was useful for experimental optimization. The results of modeling showed that this new approach could successfully predict experimental results. Under optimum conditions, an estimated particle size of about 20 nm was achieved using TEM analysis with spherical morphology. The estimated FA thickness on the surface of nanoparticles was about 5 nm. Fourier transform infrared spectroscopy (FTIR) confirmed the successful formation of FA layers around nanoparticles. Moreover, zeta potential and vibrating sample magnetometer (VSM) showed that particles had super paramagnetic property with a net charge of −52 mv, which was suitable for pharmaceutical applications.

Introduction

Benign Prostatic Hyperplasia (BPH) and prostate cancer, as common conditions in men, can be fatal in many cases. A natural medicine used for the treatment of prostate cancer is pumpkin seed oil. It is often used for treating prostate disease and other types of cancer such as breast [1], [2] and skin cancers [3] because of its unsaturated FAs content such as oleic acid (37–44%) and linoleic acid (37–44.5%) [4].

One major problem that impedes the use of pumpkin seed oil in its usual form is its low solubility in the aqueous media of body. Therefore, enhancing the solubility of pumpkin seed oil is a chief concern in its consumption. Moreover, changing the traditional form of this oil and entrapment into nanostructure can enhance its inhabitation properties. Recently, MNPs such as super paramagnetic iron oxide nanoparticles have been proposed for this purpose [5]. Coating the surface of these nanoparticles reduces their toxicity [6]. The FA coating of nanoparticles changes the zeta potential of MNPs surface to negative values, which reduces the cytotoxicity of particles [7].

There are some studies on the coating of MNPs surface with FAs. For example, oleic acid and myristic acid in the octadence solvent generate Fe₃O₄ fatty acid salt in form of nanocrystal at a temperature of 300 °C with an average particle size of 8–50 nm [8]. Li et al. [9] coated the surface of MNPs with lauric acid, citric acid and decanoic acid in aqueous environment using the hydrothermal process at 70 °C. The average estimated diameter of coated nanoparticles was about 20–30 nm. In another study, oleic acid was coated on MNPs with nanoparticles size of 10 nm in non-aqueous environment in the presence of toluene [10]. Also, ricinoleic acid in dichloromethane solvent was coated on magnetic MNPs at 60 °C, which produced an average particle size of 15 nm [11]. In addition, some studies have explored the possibility of coating salt of FAs on MNPs surfaces such as sodium oleate in an aqueous solution by autoclave at 100–230 °C, which yielded a particle size of 10.5–21 nm [12]. In all experiments, the adsorption of fatty acids on MNPs was observed in the solution inside autoclave with a constant volume solvent, which was removed after adsorption.

Furthermore, predicting optimum conditions for hydrodynamic diameter on MNPs is integral to adsorption process. Mathematical modeling could be used to achieve this goal. Neural networks including Radial Basic Function (RBF) are widely used to define experimental processes with data modeling. Regularization network (RN) technique creates a three-layer network with RBF as well as nonlinear functions without over-fitting, which provides smoothness [13]. Leave one out cross validation (LOOCV) is an RN method with cross validation [14]. Combination of RBF networks with regularization techniques creates Generalized Radial Basic Function (GRBF). There are three types of RBF nonlinear parameter estimation techniques including: three-step [15], gradient descant [16] and evolutionary algorithms [17]. In many cases, the simulated annealing (SA) algorithm optimizes RBFNN and RBF nonlinear parameters. With regard to the application of these models, the main concern is over-fitting problem, which is caused by the lack of regularization. Thus, it is necessary to find a new algorithm that can overcome the over-fitting problem.

In this study, a novel hybrid method based on SA-LOOCV-GRBF algorithm has been proposed for predicting the behavior of fatty acid adsorption on MNPs. By applying this new method, the optimum adsorption conditions for achieving the smallest nanoparticles size could be predicted. This optimum size can help achieve high-loading efficiency of FAs. More importantly, this is the first study to use pumpkin seed oil as a source of fatty acid for loading on MNPs. Another novelty of this study is related to fatty acid adsorption on MNPs. To this end, a safe and controllable process was utilized to coat FAs on MNPs in which the fatty acid adsorption on MNPs was performed during MNP drying process in the oven during the evaporation of water from solution.