Solubility Prediction in Novel Chemical Space: Combining Statistics and DFT Calculations

提出了一种改进的单/多元线性回归(S/多元线性回归)和主成分分析(PCA)等统计工具用于硅酸盐溶解度预测的方法。与常见的ClogP片断方法不同，统计分析所需的描述符将通过DFT计算生成，以涵盖溶质、溶质和溶质-溶质晶格相互作用的全局和局部结构属性。使用这些描述符的主成分分析(PCA)和化学空间映射将使选择性实验溶解度测量成为可能，以扩大该方法的化学空间覆盖范围。一组训练数据，包括难以预测的盐和离子化合物，将用于制定方案，然后用一组测试化合物进行验证和最终评估。在成功完成这些工作后，该项目将把重点转向将可靠的溶解度预测扩展到新的化学空间。

An improved protocol for in silico solubility prediction using statistical tools such assingle/multivariate linear regression (S/MLR) and principle component analysis (PCA) is proposed. Therequired descriptors for statistical analysis will be generated by DFT calculations, in contrast to the commonClogP fragments approach, to encompass both global and local structural properties of the solute, solutesolventand solute-solute lattice interactions. Principle component analysis (PCA) and chemical spacemapping with these descriptors will enable selective experimental solubility measurements to expand thechemical space coverage of the method. A training set of data, including difficult to predict salts and ioniccompounds, will be used to develop the protocol, before validation and final evaluation with a test set ofcompounds. Upon completing these successfully, the project will switch focus to the expansion of reliablesolubility prediction into novel chemical space.