Study On The QSAR For The Selected Aromatic Organic Pollutants

That the Quantitative Structure-Activity Relationship(QSAR) study of organic compounds play a very significant role in the ecological risk prediction and assessments of organic pollutants,as well as oganic pollution control and prevention.In this dissertation,the organic pollutants studied have phenyl structure,namely,are kinds of aromatic compounds,which mainly have been discharged from the industries of petrochemical,dyestuff,pesticide and pharmaceutical.Among which many(such as chlorophenol,nitrobenzene,anilinee,etc.) have been listed as priority pollutants by the U.S.EPA.In QSAR studies,both the molecular connectivity index(MCI) and quantum chemistry calculations are important way to obtain structural parameters of specific molecules.For the compounds with phenyl rings,molecular connectivity index exhibits an excellent ability for describing the structural characteristics of molecules including distinguishing the number and type of substituents,different substituent location through the point value of heteroatom and each order molecular connectivity index.In addition,quantum chemical parameters have also an explicit physical chemistry interpretation,including configuration,characteristic of covalent bond,the electronic activity,and each levels of the molecular structure.It has been reported that the most optimize QSAR model can be built by combining these two types of parametric in the studies of the activity/property of organic pollutants due to some of excellent characteristics of molecular connectivity index and quantum chemical parameters in described molecular structure information.So in this thesis,these two kinds of parameters were applicated for developing the most optimize QSAR model.Linear regression analysis is a traditional QSAR methodology;the main aim is to select the parameters in established QSAR models.The specific case studies of this thesis will be focused on the parameters selection by regression analysis in Chapter 3,4,5.Futhermore,some other methods including BP network,RBF network,and support vector machines were also adopted to develope QSAR models for estimating the activity/property of the unknown congener compounds. A brief description of QSAR development process and research status was given in Chapter 1 of this dissertation.And also expatiate on the study objective and significance of this dissertation.In Chapter 2 of this thesis,both the structural parameters and research methods in QSAR study were outlined.Including the theory definitions,calculation methods,and characteristics of each descriptor of the MCI and quantum chemical parameters were introduced in detail of the structural parameters.In the presentation of modeling methods,we stated each linear regression methods and principle comprehensive.In addition,the methods and principle of neural network and support vector machine were summarized in this chapter.In Chapter 3 of this thesis,a case study titled the QSAR studies about aromatic organic pollutants Kow and BCF was introduced.Firstly,the variable parameters such as MCI,quantum chemical parameters andΔX were selected to establish the QSAR models of Kow and BCF for 44 aromatic organic pollutants.Then the most optimize QSAR models were suggested by gradual linear regression analysis,the correlation coefficient r² were reached 0.926 and 0.914 respectively.In addition,the BP network and RBF network models were also developed by the parameters formed in the above linear regression.The study results show that the BP network model also has a well predictive ability,the correlation coefficient r² were 0.925 and 0.912 respectively.In Chapter 4,a QSAR study about Koc of aromatic organic pollutants was investigated using molecular connectivity index.In this study,the MCI of 61 aromatic compounds were calculated firstly.Then,the QSAR model between Koc of aromatic organic pollutants in soil and MCI was built by gradual linear regression method. Furthermore,¹X^v,⁴X^v_pc was choosed for the equation optimization,and the correlation coefficient r² was 0.963.In addition,the study results also show that the QSAR model of Koc by support vector machine also has a relatively predictive capacity(r² was 0.941),but not as good as the effect of the linear regression model established above.In Chapter 5 of this thesis,we discussed another study case which identified the aquatic photobacterium phosphoreum by combining MCI and quantum chemistry parameters.We firstly calculated quantum chemical descriptors of 37 aromatic organic compounds,and then constructed the QSAR model for the toxicity of aromatic compounds to aquatic photobacterium phosphoreum using quantum chemistry parameters combine ¹X^v by gradual linear regression analysis.The correlation coefficient r² is 0.933.Another significant find is that the BP networks and support vector machines also can be used to establish QSAR model for the toxicity of aromatic compounds to aquatic Photobacterium phosphoreum,the correlation coefficient r² obtained were 0.925,0.937 respectively.In Chapter 6 of this thesis,these of study contents,innovation and insufficiency overall this dissertation were summarized.Thus,a conclusion about the further development and ideas based on the summarizations is given.In this master's degree dissertation,both the molecular connectivity index(MCI) and quantum chemistry calculations were applied for the QSAR study about Kow,BCF,Koc and aquatic photobacterium phosphoreum of aromatic organic pollutants.In addition,other application methods included the gradual linear regression analysis,BP network,RBF networks and support vector machine for the most optimize QSAR models selection and development.Some results were successfully achevieved based on previous studies.It is signilicant for the aromatic organic pollutants in the prediction of activity/property using the developed QSAR models.