Analysis Of The Hydrolytic Stabilityof Halogenated Disinfection Byproducts By Use Of Quantitative Structure-activity Relationship Technique

Disinfection is the last process in water treatment, in which carcinogenic and teratogenic disinfection byproducts(DBPs) are produced. Among them, the proportion of halogenated DBPs is the largest. In pipe networks, they can experience degradation to users, in which hydrolysis is the most important. To ensure the safety of drinking water, it is essential to research the hydrolysis stability of halogenated DBPs. This paper applied quantitative structure-activity relationship to establish several models in order to predict hydrolysis property of new halogenated DBPs, which can provide basic data for the control and removal of them.For QSARs of halomethanes, the MLR model established by constitutional descriptors including induced effect descriptor σ ? and steric effect descriptor Es corresponded to the criteria of goodness-of-fit, robustness and predictivity power(R2=0.7575, Q2cv=0.7100, Q2ext=0.7878). After prediction, the higher halogenated degree was, the lower hydrolysis rate constant was. The hydrolysis rate constant is from 1.07×10-3 h-1~2.14×10-3 h-1 and the half life of all predicted halomethanes was beyond residence time.For QSARs of several types of halogenated DBPs, the MLR model built up by σ ? and Es which were both recalculated with functional groups effect were unacceptable, which meant constitutional descriptors were not applicable to research QSARs of halogenated DBPs. However, the MLR model(R2=0.8610, Q2cv=0.8245, Q2ext=0.6607) and PLS model(R2=0.8765, Q2cv=0.7950, Q2ext=0.8180) established by 30 unconventional descriptors could give good performance. Due to missing some information, PCR was not as good as MLR and PLS, and did not reach the criterion of external validat ion. ANN had no obvious advantage, either. After prediction, The half life of the majority of DBPs was beyond residence time.Moreover, leave-one-out and K-fold(K=2, 3, 5, 10) were carried out for MLR model of hydrolysis of halogenated DBPs. Considering both of the complexity of calculation and credibility, when K=5, internal validation is the most reliable.