ABSTRACT
Nanofiltration and reverse osmosis are technologies that provide medium to high rejections of organic compounds, present as emerging contaminants in water (Schafer et al., 2003; Kimura et al., 2003b). The presence of emerging contaminants has been identified in surface water bodies, sewage treatment plant effluents, and stages of drinking water treatment plants, and even at trace-levels in finished drinking water (Kolpin et al., 2002; Heberer, 2002; Castiglioni et al., 2006). The possible effects on aquatic organisms and human health, associated with the consumption of water containing low concentrations of single compounds, have been presented in toxicology studies (Escher et al., 2005; Pomati et al., 2006; Vosges et al., 2008). The studies demonstrate that researchers do not yet understand the exact risks from decades of persistent exposure to a myriad and random combination (of low levels) of pharmaceuticals, EDCs, and other organic contaminants; hence, the long-term effects of consumption of water containing low concentrations of contaminants will remain an unanswered question for the foreseeable future. Meanwhile, water treatment facilities are implementing monitoring programs, research organisations dealing with water reuse have published reports, and studies have addressed the topic (Drewes et al., 2006; Verliefde et al., 2007). An important aspect of dealing with the problem has been the identification of compound physicochemical properties and membrane characteristics to explain transport, adsorption and removal of contaminants by different mechanisms, explicitly by size/steric exclusion, hydrophobic adsorption and partitioning, and electrostatic repulsion (Kiso et al., 2001b; Ozaki and Li, 2002; Van der Bruggen and Vandecasteele, 2002; Schafer et al., 2003; Kimura et al., 2003b; Nghiem et al., 2004; Bellona and Drewes, 2005; Xu et al., 2005). A number of articles have proposed a mechanistic understanding of the interaction between membranes and organic compounds; others have tried to apply complex models to model rejection (Cornelissen et al., 2005; Kim et al., 2007; Verliefde et al., 2008). However, there have been few models to “predict” the rejection of compounds. To overcome this situation, our objective was to create a general quantitative structure-activity relationship (QSAR) model to predict rejection based on an integral approach that considers membrane characteristics, filtration operating conditions and physicochemical compound properties. A QSAR is a method that relates an activity of a set of compounds quantitatively to chemical descriptors (structure or property) of those compounds (Sawyer et
al., 2003). QSAR’s objective is predicting but maintaining a relationship to mechanistic interpretation. Applications of QSAR for the development of models to find relationships between membranes and organic compounds have been presented in journals related to drug discovery and medicinal chemistry for analysis of permeability of membranes to organic compounds (Ren et al., 1996; Fujikawa et al., 2007). The study of reverse osmosis membranes has also experienced the application of QSAR principles. Campbell et al. (1999) performed a QSAR analysis of surfactants influencing the attachment of a mycobacterium to cellulose and aromatic polyamide reverse osmosis membranes. Their objective was to understand the relationship between surfactant molecular properties and activity on the membrane surface that inhibits bacterial attachment to the membrane to reduce biofilm formation and to increase permeate production. The present study uses the concept of QSAR analysis to quantify an activity, compound rejection by a membrane, in terms of organic compound physicochemical properties, membrane characteristics (salt rejection, pure water permeability, molecular weight cut-off, charge, hydrophobicity) and operating conditions (pressure, flux, cross flow velocity, back diffusion mass transfer coefficient, recovery). In this work a QSAR model was constructed with internal experimental data used for training. The model was internally validated using measures of goodness of fit and prediction. Subsequently, after identification of a relationship in the form of an equation, estimations of rejection for an external dataset for different compounds and membranes were used to externally validate the model. Similarly, rejections of more emerging organic contaminants can be predicted in advance, before nanofiltration or reverse osmosis applications. Nevertheless, the QSAR model is applicable over the range of boundary experimental conditions that will be defined in the experimental section.
