The prediction of chlorine consumption of organic molecules in drinking water treatment

Abstract

In the present study, the chlorine demand of a diverse set of organic chemicals present in water bodies was investigated by a quantitative structure-property relationship (QSPR) model. The descriptors required for the model development were obtained by SPARTAN (v.10), DRAGON (v.6.0), and ADMET (v.8.0) software packages. The selection of descriptors was carried out via the tools implemented in QSARINS (v.2.2.1) software. Numerous division trials were performed on the data set as training and test sets which comprise the 80% and 20% of the whole data set, respectively. The generated models were validated internally and externally in line with the Organization of Economic Co-operation and Development (OECD) principles. Six descriptors from DRAGON (v.6.0) and one descriptor from ADMET (v.8.0) constitute the final model. These descriptors stem from various blocks including GETAWAY, WHIM, information indices, molecular properties, simple constitutional, 2D autocorrelation and 2D atom pairs. The predictive ability of the final model was tested using an external data set consisting of various pharmaceuticals and personal care products (PPCPs) with no experimental chlorine demand data. The proposed QSPR model covers structurally 91% of the external chemicals. The AD of the generated model was also strictly defined by the range of descriptors’ approach. The predictive ability of the generated model was found to be reliable for most of the tested PPCPs. Antibiotics are the highlighted pharmaceuticals among the tested PPCPs with the highest chlorine demand.