重新思考复合:课程学习重塑基于变压器的小分子反应预测
Rahul Sheshanarayana1, Fengqi You1,2,3,4
1College of Engineering, Cornell University, Ithaca, New York 14853, United States.
Journal of chemical information and modeling
|September 26, 2025
概括
我们开发了一个课程学习 (CL) 框架,通过训练模型对日益增加的难度的反应来改进反合成预测. 这种方法增强了化学概括,特别是复杂的反应,导致了显著的准确性收益.
科学领域:
- 计算化学计算化学
- 人工智能的人工智能
- 化学合成 化学合成
背景情况:
- 逆合成预测对于药物发现和材料科学至关重要.
- 目前的模型难以将其推广到罕见或复杂的化学反应.
- 统一的培训方法无法捕捉化学转换的细微差别.
研究的目的:
- 引入一个课程学习 (CL) 框架,以提高回复合成预测.
- 在模型训练过程中系统地控制反应难度,以改善化学概括.
- 在遇到罕见或代表性不足的反应的模型中解决故障模式.
主要方法:
- 开发了一个难度意识的课程学习框架.
- 在基于合成可访问性,环复杂性和分子大小的化学信息进展中引入反应.
- 将 CL 框架应用于基于变压器的架构 (ChemBERTa + DistilGPT2, ReactionT5v2, BART).
主要成果:
- 在所有测试的架构中获得了实质性的性能提升.
- 在没有化学预训练的情况下,CL将BART模型的top-1精度从27.0%提高到75.9%.
- 在低数据制度和基于支架/结构不相似的分割下观察到显著的改进.
- 在没有辅助标签,模板或反应类监督的情况下获得了收益.
结论:
- 课程学习框架有效地提高了回复合成预测的准确性和稳定性.
- 这种方法为规划各种化学化合物的合成路径提供了一个有希望的方法.
- 该框架显示了在制药中间体,催化剂,聚合物和功能材料中的应用潜力.
相关概念视频
Predicting Reaction Outcomes
10.1K
Kinetics describes the rate and path by which a reaction occurs. In contrast, thermodynamics deals with state functions and describes the properties, behavior, and components of a system. It is not concerned with the path taken by the process and cannot address the rate at which a reaction occurs. Although it does provide information about what can happen during a reaction process, it does not describe the detailed steps of what appears on an atomic or a molecular level. On the other hand,...
10.1K
Cycloaddition Reactions: Overview
3.4K
Cycloadditions are one of the most valuable and effective synthesis routes to form cyclic compounds. These are concerted pericyclic reactions between two unsaturated compounds resulting in a cyclic product with two new σ bonds formed at the expense of π bonds. The [4 + 2] cycloaddition, known as the Diels–Alder reaction, is the most common. The other example is a [2 + 2] cycloaddition.
3.4K
Woodward–Hoffmann Selection Rules and Microscopic Reversibility
3.8K
Electrocyclic reactions, cycloadditions, and sigmatropic rearrangements are concerted pericyclic reactions that proceed via a cyclic transition state. These reactions are stereospecific and regioselective. The stereochemistry of the products depends on the symmetry characteristics of the interacting orbitals and the reaction conditions. Accordingly, pericyclic reactions are classified as either symmetry-allowed or symmetry-forbidden. Woodward and Hoffmann presented the selection criteria for...
3.8K
Synthesis and Decomposition Reactions
38.0K
Synthesis and decomposition are two types of redox reactions. Synthesis means to make something, whereas decomposition means to break something. The reactions are accompanied by chemical and energy changes.
38.0K
SN2 Reaction: Stereochemistry
11.6K
In an SN2 reaction, the nucleophilic attack on the substrate and departure of the leaving group occurs simultaneously through a transition state. As the nucleophile approaches the substrate from the back-side, the configuration of the substrate carbon changes from tetrahedral to trigonal bipyramidal and then back to tetrahedral, leading to an inversion in the configuration of the product.
If the substrate is an achiral molecule at the α-carbon, the inversion of configuration is not...
If the substrate is an achiral molecule at the α-carbon, the inversion of configuration is not...
11.6K
Thermal and Photochemical Electrocyclic Reactions: Overview
3.0K
Electrocyclic reactions are reversible reactions. They involve an intramolecular cyclization or ring-opening of a conjugated polyene. Shown below are two examples of electrocyclic reactions. In the first reaction, the formation of the cyclic product is favored. In contrast, in the second reaction, ring-opening is favored due to the high ring strain associated with cyclobutene formation.
3.0K


