Q-DFTNet:一种基于化学的神经网络框架,可以通过DFT驱动的QM9数据来预测分子双极时刻
Dennis Delali Kwesi Wayo1, Mohd Zulkifli Bin Mohamad Noor1, Masoud Darvish Ganji2
1Faculty of Chemical and Process Engineering Technology, Universiti Malaysia Pahang Al-Sultan Abdullah, Kuantan, Malaysia.
Journal of computational chemistry
|August 13, 2025
概括
图形神经网络 (GNN) GraphConv 证明了在预测分子二极点时刻方面具有最佳的准确性和效率. 这种化学信息神经网络 (ChINN) 框架,Q-DFTNet,为量子化学应用提供了强大的基线.
科学领域:
- 计算化学计算化学
- 机器学习 机器学习
- 量子力学就是量子力学.
背景情况:
- 在计算化学和材料科学中,预测双极时刻等分子性质至关重要.
- 图形神经网络 (GNN) 对分子性质预测有希望,但需要仔细的基准测试.
- 对于大规模应用,现有的模型可能缺乏可解释性或计算效率.
研究的目的:
- 介绍Q-DFTNet,一个化学信息神经网络 (ChINN) 框架用于基因基因网络的基准测试.
- 在QM9数据集上评估七个GNN架构用于二极极矩预测.
- 确定提供精度,可解释性和计算效率的最佳平衡的GNN.
主要方法:
- 在QM9数据集上训练了七个GNN架构 (GCN,GIN,GraphConv,GATConv,GATNet,SAGEConv,GIN+EdgeConv) 在100个时代.
- 使用性能指标 (MSE,MAE,R^2) 和可解释性分析 (t-SNE,PCA,UMAP,残余图) 评估的模型.
- 根据可训练的参数和预测性表现评估准确性-复杂性权衡.
主要成果:
- 在最小的参数 (16.5k) 中,GraphConv获得了最低的测试MSE (0.7054) 和MAE (0.6196),表明精度-复杂性优越.
- GIN+EdgeConv表现出强的表现,利用边缘意识提高表现力.
- 基于注意力的模型 (GATConv,GATNet) 尽管复杂性更高,但表现不佳,而隐性空间分析显示GraphConv,GIN+EdgeConv和GCN的集群分离性更好.
结论:
- 在Q-DFTNet框架内,GraphConv为双极时刻预测提供了最佳的精度-复杂性权衡.
- Q-DFTNet为GNN在量子化学和材料发现中的部署提供了化学基础的基线.
- 该框架强调了模型架构选择对于实现可解释和高效的分子性质预测的重要性.
相关概念视频
Molecular Geometry and Dipole Moments
14.0K
The VSEPR theory can be used to determine the electron pair geometries and molecular structures as follows:
14.0K
Electric Dipoles and Dipole Moment
5.5K
Consider two charges of equal magnitude but opposite signs. If they cannot be separated by an external electric field, the system is called a permanent dipole. For example, the water molecule is a dipole, making it a good solvent.
Theoretically, studying electric dipoles leads to understanding why the resultant electric forces around us are weak. Since electric forces are strong, remnant net charges are rare. Hence, the interaction between dipoles helps us understand electrical interactions in...
Theoretically, studying electric dipoles leads to understanding why the resultant electric forces around us are weak. Since electric forces are strong, remnant net charges are rare. Hence, the interaction between dipoles helps us understand electrical interactions in...
5.5K
Induced Electric Dipoles
4.4K
A permanent electric dipole orients itself along an external electric field. This rotation can be quantified by defining the potential energy because the external torque does work in rotating it. Then, the potential energy is minimum at the parallel configuration and maximum at the antiparallel configuration. While the former is a stable equilibrium, the latter is an unstable equilibrium.
Since the absolute value of potential energy holds no physical meaning, its zero value can be chosen as per...
Since the absolute value of potential energy holds no physical meaning, its zero value can be chosen as per...
4.4K
Bond Polarity, Dipole Moment, and Percent Ionic Character
30.1K
Bond Polarity
30.1K
¹³C NMR: Distortionless Enhancement by Polarization Transfer (DEPT)
1.2K
When proton-coupled carbon-13 spectra are simplified by a broadband proton decoupling technique, structural information about the coupled protons is lost. Distortionless enhancement by polarization transfer (DEPT) is a technique that provides information on the number of hydrogens attached to each carbon in a molecule. While the DEPT experiment utilizes complex pulse sequences, the pulse delay and flip angle are specifically manipulated. The resulting signals have different phases depending on...
1.2K
Molecular Shape and Polarity
62.0K
Dipole Moment of a Molecule
62.0K


