SMG-BERT:集成立体信息和化学表示,用于分子性质预测
Jiahui Zhang1,2, Wenjie Du1,2, Xiaoting Yang2,3
1School of Software Engineering, University of Science and Technology of China, Hefei, China.
Frontiers in molecular biosciences
|July 17, 2023
概括
立体分子图 BERT (SMG-BERT) 通过整合1D,2D和3D信息来增强分子性质预测. 这种机器学习方法准确地预测了属性并区分了立体异构体,改进了现有的方法.
科学领域:
- 计算化学计算化学
- 机器学习 机器学习
- 药物发现 药物发现 药物发现
背景情况:
- 准确的分子性质预测在科学学科中至关重要.
- 传统方法和现有的机器学习 (ML) 模型与立体异构体差异化作斗争,缺乏可解释的表示.
- 当前的ML模型往往无法包含关键的3D分子信息.
研究的目的:
- 开发一种新的机器学习模型,用于准确和可解释的分子性质预测.
- 改进立体异构体的分化,特别是性分子.
- 使用多维化学信息建立统一的分子表征系统.
主要方法:
- 拟议的立体分子图形BERT (SMG-BERT),集成3D几何,2D拓和1D SMILES字符串数据.
- 使用基于自我注意的BERT架构.
- 包含核磁共振 (NMR) 光谱和键解离能 (BDE) 作为额外的原子和键特性.
主要成果:
- 在12个基准分子数据集中,SMG-BERT表现出卓越的性能.
- 该模型有效地区分了形状和性分子.
- 实现了与化学原理相一致的增强模型解释性.
结论:
- 通过利用全面的1D,2D和3D信息,SMG-BERT在分子性质预测方面取得了重大进展.
- 该模型能够整合多样化的化学数据,提高了准确性和可解释性.
- SMG-BERT被证明是计算化学和相关领域的可通用和可靠工具.
相关概念视频
Molecular Models
38.7K
Physical models representing molecular architectures of chemical compounds play essential roles in understanding chemistry. The use of molecular models makes it easier to visualize the structures and shapes of atoms and molecules.
38.7K
Predicting Molecular Geometry
34.5K
VSEPR Theory for Determination of Electron Pair Geometries
34.5K
Fischer Projections
13.4K
Learning to draw Fischer projections of molecules and understanding their relevance plays a crucial role in the visual depiction of organic molecules. A Fischer projection is a two-dimensional projection on a planar surface to simplify the three-dimensional wedge–dash representation of molecules. This is especially helpful in the case of molecules with multiple chiral centers that can be difficult to draw. Here, all the bonds of interest are represented as horizontal or vertical lines.
13.4K
Polymer Classification: Stereospecificity
2.5K
Polymerization generates chiral centers along the entire backbone of a polymer chain. Accordingly, the stereochemistry of the substituent group has a significant effect on polymer properties. Polymers formed from monosubstituted alkene monomers feature chiral carbons at every alternate position in the polymer backbone. Relative to the predominant orientation of substituents at the adjacent chiral carbons, the polymer can exist in three different configurations: isotactic, syndiotactic, and...
2.5K
Stereoisomers
13.1K
On the basis of mirror symmetry, stereoisomers of an organic molecule can be further classified into diastereomers and enantiomers. Diastereomers are stereoisomers that are not mirror images of each other. Substituted alkenes, such as the cis and trans isomers of 2-butene, are diastereomers, as these molecules exhibit different spatial orientations of their constituent atoms, are not mirror images of each other, and do not interconvert. Here, the interconversion is suppressed due to...
13.1K
Stereoisomerism of Cyclic Compounds
9.0K
In this lesson, we delve into the role of ring conformation and its stability, which determines the spatial arrangement and, consequently, the molecular symmetry and stereoisomerism of cyclic compounds. 1,2-Dimethylcyclohexane is used as a case study to evaluate the possible number of stereoisomers. Here, given the multiple (n = 2) chiral centers, there are 2n = 4 possible configurations that lack a plane of symmetry, as the ring skeleton exists in a non-planar chair conformation. In addition,...
9.0K


