Dataset for the evaluation of Supercritical Fluid Chromatography Non-Polar Stationary Phases: HSS C18 SB, CSH PFP, PGC

Raw data used for the evaluation of supercritical fluid chromatography non-polar stationary phases published in article "AI-Enhanced Understanding of Retention Interactions in Supercritical Fluid Chromatography: Neural Network Insights into Retention on Selected Non-Polar Stationary Phases" in Analytical Chemistry, 2025. Data set contains: (i) chromatograms of 107 analytes measured on high strenght silica C18 (HSS C18 SB), charged surface hybrid pentafluoro phenyl (CSH PFP), and porous graphitic carbon (PGC) column using methanol, 10 mmol/L ammonium in methanol, and 2% water in methanol as organic modifiers in 8 points during 1 year (Empower project, Excel sheets of retention times and measured mixtures), (ii) 226 molecular descriptors calculated by CDK Descriptor Calculator (v.1.4.8) from 3D structures of the 107 analytes optimized by semi-empirical AM1 quantum mechanical calculations using the MOPAC application of Chem 3D Pro version 14.0 software (CambridgeSoft) (Excel sheet), (iii) weights assigned to each molecular descriptor at each chromatographic conditions by artificial neural network created using the neural network simulator in Matlab R2023a with the deep learning toolbox V.23.2 (The MathWorks, Inc., Massachusetts, USA) and a sigmoid activation function, a backpropagation learning algorithm with 500 learning cycles (Excel sheet).本数据集用于评估超临界流体色谱（Supercritical Fluid Chromatography）非极性固定相，相关研究成果发表于2025年《Analytical Chemistry》期刊的论文《AI-Enhanced Understanding of Retention Interactions in Supercritical Fluid Chromatography: Neural Network Insights into Retention on Selected Non-Polar Stationary Phases》。本数据集包含以下内容：(i) 107种分析物的色谱图：在高强度硅胶C18（High Strength Silica C18, HSS C18 SB）、带电表面杂化五氟苯基（Charged Surface Hybrid Pentafluoro Phenyl, CSH PFP）以及多孔石墨碳（Porous Graphitic Carbon, PGC）色谱柱上，分别以甲醇、含10 mmol/L氨的甲醇溶液以及体积分数2%水的甲醇溶液作为有机改性剂，在1年内的8个采样节点完成数据采集，相关数据存储于包含保留时间与待测混合物信息的Empower项目Excel工作表中；(ii) 107种分析物的226个分子描述符：先通过Chem 3D Pro 14.0软件（CambridgeSoft公司）的MOPAC模块，采用半经验AM1量子力学计算对分析物的三维结构进行优化，再通过CDK描述符计算器（CDK Descriptor Calculator, v.1.4.8）计算得到分子描述符，相关数据存储于Excel工作表中；(iii) 各色谱条件下各分子描述符的权重：采用美国马萨诸塞州The MathWorks, Inc.公司的Matlab R2023a软件及其深度学习工具箱V.23.2中的神经网络模拟器构建人工神经网络，以Sigmoid激活函数与反向传播学习算法完成500次训练周期后，得到各描述符在对应色谱条件下的权重，相关数据存储于Excel工作表中。