基于QSPR方法的烃类物质苯胺点预测

Forecasting the aniline points of the hydrocarbons based on the analysis of the quantitative structure-property relationship

烃类物质在石油工业中有着非常广泛的应用。在石油工业中常用苯胺点来衡量有机溶剂的溶解性能。基于定量结构-性质相关性(QSPR)原理,根据分子结构计算反映分子结构信息的结构参数,应用遗传函数算法从大量结构参数中优化筛选出与烃类物质苯胺点最为密切相关的结构参数作为表征相应化合物结构特征的分子描述符,采用多元线性回归方法对分子描述符与苯胺点之间的定量函数关系进行关联,建立了预测烃类物质苯胺点的理论模型。最后,对模型进行了内部及外部验证来检验模型的可靠性。在此基础上,对所建立的预测模型进行机理解释,分析了影响烃类物质苯胺点的主要结构因素及其影响规律。研究表明,所建模型具有较高的稳定性和预测能力。

The paper is intended to make a prediction study of the aniline points of the hydrocarbons based on the analysis of the quantitative structure-property relationship based on the working principle of the quantitative structure-property relationship（QSPR）.As is known,it is possible to calculate the structures of the polymers of organic chemicals and reflect the characteristic features of such molecules according to the molecular structures.And,then,the genetic algorithm has been used to screen out the most appropriate molecular descriptions closely related to the aniline points of hydrocarbons.At this stage,we have chosen the multivariate linear regression method（MLR） to imitate the likely functional relationship between the selected descriptors and the aniline points.In so doing,we have obtained a theoretical model for forecasting the aniline points of hydrocarbons.To make the model as much reliable as possible,we have carefully examined and improved on the internal and external validity so as to guarantee its reliability.Specifically speaking,by the so-called ＂leave one out＂ cross validation coefficient of the prediction model we have developed for predicting the hydrocarbons is over 0.5with the multiple correlation coefficient being over 0.6,which is expected to meet the demands of successful QSPR models.What is more,we have identified the mechanism of the developed model and clarified the main factors affecting the aniline point and the changing regularities.The result of our investigation and simulation show that the main structure features affecting the aniline points of hydrocarbons can be started as follows：the molecular steric effects,including the molecular shape and degree of difference in the molecular elements and their molecular electrostatic features,so that they can have favorable impacts on the aniline points of the above-said hydrocarbons.At the same time,we have also made such aniline points of the hydrocarbons decrease with increase of the difference in molecular constituent parts.That is to say,the more complex the molecular shapes are,the higher the aniline points of hydrocarbons will be.And,in turn,the aniline points of hydrocarbons tends to increase with increase of the atomic polarizabilities of the polymers.On the other hand,the effects of the molecular atomic electronegativities on the aniline points of hydrocarbons turn to be lower than before.To be sure,they help to enhance the predictive power of the model though they do not play the major role in the structure-property relationship of the model.Thus,it can be concluded that the model has been made robust and more powerful in its prediction or forecasting power.