探索BERT用于反应收益率预测:评估代币化,分子表示和预训练数据增强的影响
Adrian Krzyzanowski1, Stephen D Pickett1, Peter Pogány1
1GSK Medicines Research Centre, Gunnels Wood Road, Stevenage SG1 2NY, U.K.
Journal of chemical information and modeling
|May 1, 2025
概括
机器学习模型可以有效地预测化学反应产量. 关键因素包括令牌化和预训练数据,对抗训练增强了合成化学应用的模型稳定性.
科学领域:
- 计算化学是一种计算化学.
- 机器学习在化学中的应用
- 合成化学 合成化学
背景情况:
- 预测反应产量在合成化学中至关重要但具有挑战性.
- 高通量实验 (HTE) 数据为模型培训提供了宝贵的信息.
- 基于BERT的模型显示出对反应产量预测的前景.
研究的目的:
- 系统地评估影响BERT基收益率预测模型的因素.
- 评估代币化,分子表示和预训练数据的影响.
- 调查对抗训练对改善模型性能的有效性.
主要方法:
- 利用基于BERT的模型来预测布赫瓦尔德-哈特维格和苏苏基-米亚乌拉合反应的产量.
- 评估了各种代币化方法 (BPE,SentencePiece,WordPiece) 和分子表示 (SMILES,DeepSMILES,SELFIES等). ) 的情况.
- 使用不同的预训练数据大小和探索人工生成的数据集,以及对抗性训练,比较性能.
主要成果:
- 分子表示的影响最小; BPE 和 SentencePiece 代币化表现最好.
- 较小的预训练数据集 (<100K反应) 与较大的数据集产生了可比的结果.
- 混合预训练集 (真实+人工数据) 和对抗训练提高了模型的稳定性和通用性.
结论:
- 代币化和预训练策略显著影响化学产量预测的ML模型性能.
- 人工生成的数据可以作为现实世界数据集的有效替代品.
- 对抗性培训提供了一种新的方法,以提高ML模型在合成化学中的稳定性和适用性.
更多相关视频
06:19Integration of Animal Behavioral Assessment and Convolutional Neural Network to Study Wasabi-Alcohol Taste-Smell Interaction
Published on: August 16, 2024
350
09:47Author Spotlight: Advancing Alzheimer's Research – Exploring Early Detection and Multi-Omics Approaches
Published on: December 15, 2023
910
相关概念视频
Predicting Reaction Outcomes
7.8K
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,...
7.8K
Reaction Quotient
47.5K
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
47.5K
Reaction Yield
48.7K
The theoretical yield of a reaction is the amount of product estimated to form based on the stoichiometry of the balanced chemical equation. The theoretical yield assumes the complete conversion of the limiting reactant into the desired product. The amount of product that is obtained by performing the reaction is called the actual yield, and it may be less than or (very rarely) equal to the theoretical yield.
48.7K
Rate-Determining Steps
31.3K
Relating Reaction Mechanisms
In a multistep reaction mechanism, one of the elementary steps progresses significantly slower than the others. This slowest step is called the rate-limiting step (or rate-determining step). A reaction cannot proceed faster than its slowest step, and hence, the rate-determining step limits the overall reaction rate.
The concept of rate-determining step can be understood from the analogy of a 4-lane freeway with a short-stretch of traffic-bottleneck caused due to...
In a multistep reaction mechanism, one of the elementary steps progresses significantly slower than the others. This slowest step is called the rate-limiting step (or rate-determining step). A reaction cannot proceed faster than its slowest step, and hence, the rate-determining step limits the overall reaction rate.
The concept of rate-determining step can be understood from the analogy of a 4-lane freeway with a short-stretch of traffic-bottleneck caused due to...
31.3K
Classification of Titrimetric Analysis Based on Reaction Types
674
Titrimetric analysis in solution chemistry involves measuring the volume of solutions and is often called volumetric analysis. The standard solution of known concentration in the burette is called the titrant, whereas the solution of unknown concentration in the flask is called the analyte, or titrand. Titrimetric analyses can be classified into four types based on the reactions between the titrant and analyte.
Titrations between an acid and a base lead to neutralization reactions that form...
Titrations between an acid and a base lead to neutralization reactions that form...
674
Multi-Step Reactions
7.2K
Chemical reactions often occur in a stepwise fashion involving two or more distinct reactions taking place in a sequence. A balanced equation indicates the reacting species and the product species, but it reveals no details about how the reaction occurs at the molecular level. The reaction mechanism (or reaction path) provides details regarding the precise, step-by-step process by which a reaction occurs. Each of the steps in a reaction mechanism is called an elementary reaction. These...
7.2K
