Designing Chemical Reaction Arrays Using Phactor and ChatGPT
Designing Chemical Reaction Arrays Using Phactor and ChatGPT
复制标题
DOI：
10.1021/acs.oprd.3c00186
复制
发表时间：
2023-08-01
影响因子：
3.4
通讯作者：
Cernak, Tim
中科院分区：
文献类型：
--
作者：
Mahjour, Babak;Hoffstadt, Jillian;Cernak, Tim
高通量实验伊萨化学合成优化的常用方法。化学家设计反应阵列来优化构件之间的偶联产量。药物研究中常用的反应包括酰胺偶联、Suzuki偶联和Buchwald-Hartwig偶联。我们展示了人工智能（AI）语言模型ChatGPT如何根据它所训练的文献记录自动为这些常见反应制定反应数组。重要的是，我们展示了ChatGPT结果如何直接转换为管理软件phactor的输入，从而实现自动执行和分析测定。实验证明了该工作流程，在第一次尝试的每种情况下都获得了中等至优异的产物产率。
High-throughput experimentation isa common practicein the optimizationof chemical synthesis. Chemists design reaction arrays to optimizethe yield of couplings between building blocks. Popular reactionsused in pharmaceutical research include the amide coupling, Suzukicoupling, and Buchwald-Hartwig coupling. We show how the artificialintelligence (AI) language model ChatGPT can automatically formulatereaction arrays for these common reactions based on the literaturecorpus it was trained on. Critically, we showcase how ChatGPT resultscan be directly translated into inputs for the management softwarephactor, which enables automated execution and analysis of assays.This workflow is experimentally demonstrated, with modest to excellentyields of products obtained in each instance on the first attempt.