负化学数据增强了反应结果预测中的语言模型
Alessandra Toniato1,2, Alain C Vaucher1,2, Teodoro Laino1,2
1IBM Research Europe, Zurich, Switzerland.
Science advances
|June 13, 2025
概括
化学反应的负结果可以改善预测模型,特别是有限的成功数据. 这种方法增强了机器学习模型的化学反应预测,即使有稀疏的积极例子.
科学领域:
- 化学 化学 化学
- 机器学习 机器学习
- 人工智能的人工智能
背景情况:
- 传统的化学研究往往会丢弃负面的实验结果,这代表了数据丰富的错失机会.
- 不充分利用负反应数据限制了预测模型的性能,特别是在数据稀缺的场景中.
研究的目的:
- 为了证明负化学反应数据在增强反应性预测模型中的有用性.
- 开发和评估一种机器学习方法,利用未成功的实验结果.
主要方法:
- 语言模型的扩展调整与强化学习用于化学反应预测.
- 训练了一个变压器模型,包括来自受控和高通量选数据集的负数据.
- 在不同的数据集上评估模型性能,积极与负面示例的比例各不相同.
主要成果:
- 在化学反应预测方面取得了最先进的性能.
- 成功地利用最小的积极数据 (仅为20分) 当得到显著更大的负面数据集 (40x) 的支持时.
- 在受控和高通量实验数据中展示了一致的模型改进.
结论:
- 负的实验结果是一个有价值的,未充分利用的资源,用于改进化学反应率预测.
- 机器学习模型,特别是基于变压器的机器学习模型,可以通过结合负面结果来有效地使用有限的积极数据进行训练.
- 这种方法为加速化学发现和优化实验设计提供了显著的优势.
相关概念视频
Predicting Reaction Outcomes
8.3K
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,...
8.3K
Reaction Quotient
48.3K
The status of a reversible reaction is conveniently assessed by evaluating its reaction quotient (Q). For a reversible reaction described by m A + n B ⇌ x C + y D, the reaction quotient is derived directly from the stoichiometry of the balanced equation as
48.3K
Predicting Products: SN1 vs. SN2
13.3K
Nucleophilic substitution reactions of alkyl halides can proceed via an SN1 or an SN2 mechanism. While in SN2 reactions, the nucleophile attacks the substrate simultaneously as the leaving group departs, in SN1 reactions, the substrate first dissociates to give the carbocation intermediate. Various factors such as the structure of the substrate, the strength of the nucleophile, and the nature of the solvent promote one mechanism over the other.
With increased substitution on the alkyl halide,...
With increased substitution on the alkyl halide,...
13.3K
2° Amines to N-Nitrosamines: Reaction with NaNO2
4.1K
Secondary amines react with nitrous acid to form N-nitrosamines, as depicted in Figure 1. Nitrous acid, a weak and unstable acid, is formed in situ from an aqueous solution of sodium nitrite and strong acids, such as hydrochloric acid or sulfuric acid, in cold conditions. In the presence of an acid, the nitrous acid gets protonated. The subsequent loss of water results in the formation of the electrophile known as nitrosonium ion.
4.1K
Leveling Effect
785
In acid-base chemistry, the leveling effect refers to the limitation imposed by the solvent on the strength of acids and bases in solution. When a base stronger than the solvent's conjugate base is used, it deprotonates the solvent until the base is entirely consumed, making it ineffective against weaker acids. Conversely, an acid stronger than the solvent's conjugate acid protonates the solvent until the acid is depleted, rendering it ineffective against weaker bases. Essentially, the...
785
Regioselectivity of Electrophilic Additions-Peroxide Effect
8.5K
In the presence of organic peroxides, the addition of hydrogen bromide to an alkene yields the isomer that is not predicted by Markovnikov’s rule. For example, the addition of hydrogen bromide to 2-methylpropene in the presence of peroxides gives 1-bromo-2-methylpropane. This addition reaction proceeds via a free radical mechanism, which reverses the regioselectivity. The free radical reaction mechanism involves three stages: initiation, propagation, and termination.
8.5K


