分布构效关系方法及其在计算热力学性质中的应用

The distributive structure-property/activity relationship method will be proposed in this project, which would be further universally used for accurate description and prediction of various thermodynamics properties, such as critical properties, the vaporization enthalpy, the fusion enthalpy, octanol-water partition coefficient, chromatographic retention and the toxicity/activity of organic compounds. Basically, this method is consisted of two parts, the norm descriptor of the molecule character matrix and the distributive structure-activity relationship function. Here, this norm descriptor of the molecule character matrix is composed of two parts: the Euclidean spatial distance matrix and the atom character property identify matrix of the molecule, by which the specialty point could be avoided basically and/or made up thoroughly. Further, this distributive structure-property/activity relationship function is developed based on our previous general distributive group contribution function. Accordingly, this new method could be universal and used for multi-objectives including thermodynamics properties/ activities/toxicities. In addition, this new distributive structure-property/activity relationship method could be used not only for macromolecules and heterocyclic compounds of organics, but also for drug molecular system. Consequently, using this new method, some compounds with novel type structures might be deduced and produced based on the calculation results. This means that the lead/precursor compound with designated performance and character could be induced and generated just through the direct calculation process. Also, in order to insure the stability and accuracy of the prediction results, nonlinear Kalman filter is to be used effectively for the model development, for which the calculation target is to find “the true value” of physi-chemical properties but not “the smallest residuals summation” as the traditional physi-chemical properties estimation methods did.

本项目拟提出分布构效关系方法.该方法将以普遍化方式对有机物临界性质、蒸发焓、熔化焓、辛醇/水分配系数、色谱保留等热力学性质以及有机物活性/毒性进行准确描述和预测.分布构效关系方法由两部分组成:分子特征矩阵范数描述符和分布构效关系函数.分子特征矩阵范数描述符由原子在分子中欧氏空间距离矩阵和特征属性识别矩阵两部分构成,对其所描述的对象避免了特异性指向;分布构效关系函数建立在广义分布贡献函数基础上,对有机物的多目标(热力学性质/活性/毒性等)描述具有广义性指向.分布构效关系方法的提出,目的在于对目标体系描述不再受制于同系物或同源物的束缚,即不仅对有机物大分子及杂环分子通用，还适用于大分子药物体系.运用本方法计算结果可能直接产生新的结构类型,即通过直接计算、设计产生具有指定性能和特征的材料先导物.数据处理拟采用非线性Kalman滤波器,以寻找数据真值为目标,确保预测目标性质数据的稳定性和准确性.

物性预测方法在现代化学化工、生物化学、材料设计、药物设计和环境保护等领域发挥着至关重要的作用。.本项目旨在研究大分子及多重杂环有机物的组成、结构与性能之间的关系，核心是建立新的分布构效关系方法。该方法将以普遍化方式对有机物临界性质、蒸发焓、熔化焓和辛醇/水分配系数等热力学性质及有机物的毒性与活性进行准确描述和预测。.本项目建立分子特征矩阵范数系列描述符，新定义的分子特征矩阵范数描述符包含两部分，欧氏空间距离矩阵和性质矩阵。距离矩阵是依据分子优化结果获得的各个原子在分子中的空间位置而建立起来的矩阵。性质矩阵是对应距离矩阵建立起来的针对各个原子的特征识别矩阵及其帽子矩阵。分子特征矩阵范数描述符不仅能够定量描述化合物分子的原子组成、连接方式，而且对原子的空间结构、空间构象也能准确反映，能很好地区分包括顺反异构等同分异构体。据此，本项目建立分布构效关系数学模型，普遍化方式对有机物热力学性质及毒性与活性进行准确描述和预测。.研究结果表明，统一的分子特征矩阵范数系列描述符，能够准确描述有机物临界性质、沸点、蒸发焓、熔化焓、辛醇/水分配系数以及有机物的毒性与活性等性质；统一的定位分布构效关系函数能够对有机物的临界性质，标态下沸点，熔点，蒸发焓（正常沸点下）和闪点等多种热力学性质和有机物的毒性活性进行准确计算。与文献方法对比，本研究计算结果无论在准确性上，还是通用性方面，都有显著改善。同时，本研究结果还表明，本项目提出建立的“定位分布构效关系函数”数学模型具有统一性和普适性。