gSelformer-MV:多视图,子图增强组SELFIES转换器用于分子性质预测
Vadim Korolev1,2, Alexey Andreevich Sorokin1,2, Yuri Kuratov2
1AI Center, Lomonosov Moscow State University, Moscow 119991, Russia.
Journal of chemical information and modeling
|December 23, 2025
概括
我们开发了gSelformer-MV,这是一个新的变压器模型,集成功能组信息用于分子性质预测. 与使用SELFIES字符串的现有方法相比,这种方法提高了准确性和可解释性.
科学领域:
- 计算化学计算化学
- 机器学习 机器学习
- 化学信息学 化学信息学
背景情况:
- 数据驱动的方法对于将化学结构与特性联系起来至关重要.
- 目前的分子表示学习主要使用以原子为中心的方法 (例如,图形神经网络,化学语言模型).
- 将功能组信息集成到先进模型中是一个尚未探索的领域.
研究的目的:
- 介绍gSelformer-MV,一个设计用于在原子和亚结构层次上表示分子的变压器模型.
- 为了利用组SELFIES的多个视图 (增强功能组令牌的SELFIES的变体) 来改进分子性质预测.
- 解决将功能组数据纳入高级分子表示学习中的差距.
主要方法:
- 开发了gSelformer-MV,这是一个在集团SELFIES的多个视图上运行的变压器架构.
- 构建多个子图分区集团SELFIES视图用于联合培训和推理.
- 将gSelformer-MV与仅在SELFIES字符串上训练的模型进行比较.
主要成果:
- gSelformer-MV表现出比仅使用SELFIES的模型更高的准确性和可解释性.
- 在多个分子回归基准上实现了最先进的性能.
- 通过专注于高可信度预测,观察到进一步的性能改善.
结论:
- 使用组 SELFIES 的子图增大是增强基于字符串的分子性质预测的有效策略.
- gSelformer-MV为分子表示学习提供了一种强大的新方法.
- 这些发现突显了将子结构信息纳入化学预测建模的潜力.
相关概念视频
02:27Predicting Molecular Geometry
44.4K
VSEPR Theory for Determination of Electron Pair Geometries
44.4K
02:00Molecular Models
43.3K
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.
43.3K
01:09Multi-pass Transmembrane Proteins and β-barrels
6.3K
In multi-pass transmembrane proteins, the polypeptide chain crosses the membrane more than once. The transmembrane polypeptide chain either forms an α-helix or β-strand structure. α-Helix containing multi-pass transmembrane proteins are ubiquitous, whereas β-strand containing ones are mainly found in gram-negative bacteria, mitochondria, and chloroplasts.
α-Helix containing multi-pass transmembrane proteins
Multi-pass transmembrane proteins such as...
α-Helix containing multi-pass transmembrane proteins
Multi-pass transmembrane proteins such as...
6.3K
02:36Molecular Geometry and Dipole Moments
17.7K
The VSEPR theory can be used to determine the electron pair geometries and molecular structures as follows:
17.7K
01:07Electron Microscope Tomography and Single-particle Reconstruction
2.8K
Transmission electron microscopy (TEM) can be used to determine the 3D structure of biological samples with the help of techniques such as electron microscope tomography and single-particle reconstruction. While single-particle reconstruction can examine macromolecules and macromolecular complexes in vitro conditions only, tomography permits the study of cell components or small cells in vivo.
Electron Tomography
Electron tomography can be performed either in TEM or STEM (scanning transmission...
Electron Tomography
Electron tomography can be performed either in TEM or STEM (scanning transmission...
2.8K
00:56Extraction: Advanced Methods
1.0K
Metal ions can be separated from one another by complexation with organic ligands–the chelating agent– to form uncharged chelates. Here, the chelating agent must contain hydrophobic groups and behave as a weak acid, losing a proton to bind with the metal. Since most organic ligands used in this process are insoluble or undergo oxidation in the aqueous phase, the chelating agent is initially added to the organic phase and extracted into the aqueous phase. The metal-ligand complex is...
1.0K