Quantitative Structure-Activity Relationship (QSAR) studies are proposed to establish the quantitative correlations between the toxicity of disinfection by-products (DBPs) and their molecular structure descriptors such as log P, Hammet sigma o, and molecular connectivity (MC) indices. The major toxicity database to be used is Computox Toxicity Database (version 5.0) developed by Dr. Kaiser, the consultant of this project. In addition, other toxicity databases and newly reported toxicity results will be extensively collected. Comprehensive databases of toxicity and molecular descriptors of DBPs would be constructed according to major chemical classes such as halogenated alkane, alkene, aromatic, aldehyde, ketone, carboxylic acid, and heterocycle. About 500 DBPs will be used in the training set for the development of QSAR models. The Hansch QSAR method which can probe the underlining modes of action for each correlation model will be used as one of the correlation models. Since MC is one of the most widely used molecular descriptors and is based upon molecular skeleton, electron, and orbital counts encoding information about topological, geometric, spatial, and electronic attributes of the chemicals. Different levels of MC indices will be used to replace the Taft steric constant E, in the Hansch analysis. QSAR models will be developed for the Microtox toxicity data. According to the developed QSAR models, four modes of action, namely, nonpolar narcotic, polar narcotic, uncoupling, and reactive toxicity, will be assigned to each corresponding QSAR model depending upon the contribution of each molecular descriptor to the overall toxicity. The same QSAR analysis will be conducted for the toxicities of DBPs to different fish, rats, and mice. The resulted QSAR models will make it possible to correlate the inter-species toxicities using the Microtox toxicity QSAR models as the baseline. The modest action would also be discussed for each QSAR equation. Computer softwares such as CFUNC would be used to calculate the MC indices for the 500 different DBPs in the training set and about 500 other potential DBPs. In order to establish the QSAR models of a set of toxic data of DBPs and their corresponding MC indices, statistic softwares such as MINITAB and SAS will be used to optimize variables, to select models, and to conduct regression and coefficient analysis. The developed QSAR models would have three important applications. First, they can be used to quantitatively predict the toxicity of any given DBP so that the regulatory agency such as the EPA can prioritize the toxic potency of a given DBP with respect to the others. Second, since different levels of MC indices encode different structural information, reactive functional groups or fragments can be inferred for major toxic mechanisms according to the contribution of MC indices to the toxicity. Third, a computer software containing the QSAR models and the corresponding toxicological mechanisms involved in each class of DBPs will be developed. The software can be attached to the EPA's database of DBPs, directly to aid the EPA regulatory activities, including DBPs priority screening, toxicity data prediction, re-examination and recommendation of the Maximum Contamination Levels of DBPs.
