ANI-aa-qmmm Data set

This data set was constructed for our work entitled "Assessment of embedding schemes in a hybrid machine learning/classical potentials (ML/MM) approach" to provide an unbiased insight into the performance of different atomic partial charges and embedding schemes to reach the <i>state of the art</i> electrostatic scheme level in multiscale frameworks.This data set was constructed using all the amino acid isomer species of the ANI-1x and ANI-2x data sets (only sulfur-containing species of the last one), which corresponded to about 80k structures <i>in vacuo</i>. We complemented the previously computed properties with new sets of atomic partial charges, energies, atom types, and QM/MM properties corresponding the structures solvated in TIP3P water molecules.A README.txt file to navigate the data set is provided.本数据集专为我们题为《混合机器学习/经典势能（ML/MM）方法中的嵌入方案评估》的研究构建，旨在为客观分析不同原子部分电荷与嵌入方案的性能提供支撑，以在多尺度框架中达成当前最优水准（state of the art）的静电方案水平。本数据集采用ANI-1x与ANI-2x数据集的全部氨基酸异构体物种构建而成（仅保留后者中的含硫物种），涵盖约8万个真空环境下（in vacuo）的分子结构。我们还为此前计算得到的属性补充了新的数据集，包含与TIP3P水分子溶剂化结构对应的原子部分电荷、能量、原子类型以及量子力学/分子力学（QM/MM）属性。本数据集附带README.txt文件，用于指导用户浏览与使用。