Supplementary Material for the article "Ready-to-use, simple, and accurate empirical equations for the estimation of gas chromatographic retention indices for the DB-225MS stationary phase"

The models published in the literature for predicting the value of the gas chromatographic retention index based on the structure of a molecule are not frequently employed in practice due to their low accuracy, the necessity for the use of paid software to calculate molecular descriptors, and the very narrow applicability domain of many models. In recent years, models based on deep learning have emerged that can achieve acceptable prediction accuracy for a wide range of molecules. These models are now being used in practice as an additional criterion in GC-MS identification. The DB-225MS stationary phase (50%-cyanopropylphenyl-methylpolysiloxane) is widely used in practice, but usable models for estimating retention indices are not available for this stationary phase. This study presents such models for the first time. Spreadsheet contains 6 sheets. The contents of the sheets are shown below. The file contains datasets (with sources), molecular descriptor values and prediction equations ready for use.<br>S1. Training set: compound structures, experimental retention indices (DB-225MS), predicted retention indices, molecular descriptor values S2. Test set: compound structures, experimental retention indices (DB-225MS), predicted retention indices, molecular descriptor values S3. External test set (esters)S4. External test set (phenols)S5. Information about molecular descriptorsS6. Python code snippet for calculating molecular descriptors and predicting retention indices, equations for predicting retention indices.Sources from which data are used:Yan J, Liu X-B, Zhu W-W, et al. Retention Indices for Identification of Aroma Compounds by GC: Development and Application of a Retention Index Database. Chromatographia. 2015;78(1–2):89–108. doi: 10.1007/s10337-014-2801-y.Ashes JR, Haken JK. Gas chromatography of homologous esters. Journal of Chromatography A. 1974;101(1):103–123. doi: 10.1016/S0021-9673(01)94737-5.Grzybowski J, Lamparczyk H, Nasal A, et al. Relationship between the retention indices of phenols on polar and non-polar stationary phases. Journal of Chromatography A. 1980;196(2):217–223. doi: 10.1016/S0021-9673(00)80441-0.<br><br>

以往基于分子结构预测气相色谱保留指数的文献报道模型，因精度偏低、需使用付费软件计算分子描述符（molecular descriptor），且多数模型适用范围极窄，实际应用中并未得到广泛采用。近年来，基于深度学习（deep learning）的模型应运而生，可对多种分子实现可接受的预测精度，目前已作为气相色谱-质谱联用（Gas Chromatography-Mass Spectrometry，简称GC-MS）鉴定的辅助判据投入实际使用。DB-225MS固定相（50%-氰丙基苯基-甲基聚硅氧烷）在实际中应用广泛，但目前尚无适用于该固定相的可用保留指数预测模型。本研究首次构建了此类预测模型。本数据集包含6个工作表，各表内容如下：该文件内含带来源标注的数据集、分子描述符数值以及可直接使用的预测方程。
S1. 训练集：化合物结构、实验保留指数（DB-225MS）、预测保留指数、分子描述符数值
S2. 测试集：化合物结构、实验保留指数（DB-225MS）、预测保留指数、分子描述符数值
S3. 外部测试集（酯类）
S4. 外部测试集（酚类）
S5. 分子描述符相关信息
S6. 用于计算分子描述符与预测保留指数的Python代码片段，以及保留指数预测方程。

本数据集所用数据来源如下：
严杰，刘新波，朱伟伟，等. 气相色谱鉴定香气化合物的保留指数：保留指数数据库的构建与应用. 色谱学（Chromatographia）. 2015;78(1–2):89–108. doi: 10.1007/s10337-014-2801-y.
阿什斯 JR，哈肯 JK. 同系酯类的气相色谱分析. 色谱A期刊（Journal of Chromatography A）. 1974;101(1):103–123. doi: 10.1016/S0021-9673(01)94737-5.
格日博夫斯基 J，兰帕奇克 H，纳萨尔 A，等. 极性与非极性固定相上酚类化合物的保留指数相关性. 色谱A期刊（Journal of Chromatography A）. 1980;196(2):217–223. doi: 10.1016/S0021-9673(00)80441-0.