作为计算化学中的一项基本任务，逆向合成预测旨在识别一组反应物来合成目标分子。现有的无模板方法只考虑目标分子的图结构，通常不能很好地推广到稀有反应类型和大分子。 在这里，我们提出了T-Rex，这是一种文本辅助逆转录合成预测方法，它利用预先训练的文本语言模型，如ChatGPT，来帮助生成反应物。 T-Rex首先利用ChatGPT生成目标分子的描述，并根据描述和分子图对候选反应中心进行排名。 然后，它通过查询每种反应物的描述对这些候选者进行重新排序，并检查哪组反应物最能合成目标分子。 我们观察到，T-Rex在两个数据集上的表现明显优于基于图的最先进方法，这表明了考虑文本信息的有效性。 我们进一步发现，T-Rex的表现优于仅使用基于ChatGPT的描述而没有重新排序步骤的变体，这表明我们的框架如何优于ChatGPT和图形信息的直接集成。 总的来说，我们表明，由预先训练的语言模型生成的文本可以显著改善逆向合成预测，为利用ChatGPT来推进计算化学开辟了新的途径。

• 共同一作：刘益枫（清华大学交叉信息研究院，曾于华盛顿大学春研）、许涵文（华盛顿大学计算机科学与工程学院）
  
• 通讯作者：王晟教授（华盛顿大学计算机科学与工程学院）
  
• 其他作者：方汤骐（清华大学交叉信息研究院，曾于华盛顿大学春研）、席浩诚（清华大学交叉信息研究院，曾于华盛顿大学春研）、刘梓轩（华盛顿大学计算机科学与工程学院）、Sheng Zhang（微软首席研究员）, Hoifung Poon（微软Health Futures总经理）

As a fundamental task in computational chemistry, retrosynthesis prediction aims to identify a set of reactants to synthesize a target molecule. Existing template-free approaches only consider the graph structures of the target molecule, which often cannot generalize well to rare reaction types and large molecules. Here, we propose T-Rex, a text-assisted retrosynthesis prediction approach that exploits pre-trained text language models, such as ChatGPT, to assist the generation of reactants. T-Rex first exploits ChatGPT to generate a description for the target molecule and rank candidate reaction centers based both the description and the molecular graph. It then re-ranks these candidates by querying the descriptions for each reactants and examines which group of reactants can best synthesize the target molecule. We observed that T-Rex substantially outperformed graph-based state-of-the-art approaches on two datasets, indicating the effectiveness of considering text information. We further found that T-Rex outperformed the variant that only use ChatGPT-based description without the re-ranking step, demonstrate how our framework outperformed a straightforward integration of ChatGPT and graph information. Collectively, we show that text generated by pre-trained language models can substantially improve retrosynthesis prediction, opening up new avenues for exploiting ChatGPT to advance computational chemistry.

• Co-first authors: Yifeng Liu (IIIS of Tsinghua University, previously research at University of Washington), Hanwen Xu(Paul G. Allen School of Computer Science & Engineering, University of Washington)
• Corresponding authors: Sheng Wang (Paul G. Allen School of Computer Science & Engineering, University of Washington)
• Other authors: Tangqi Fang(IIIS of Tsinghua University, previously research at University of Washington), THaocheng Xi(IIIS of Tsinghua University, previously research at University of Washington), Zixuan Liu (Paul G. Allen School of Computer Science & Engineering, University of Washington), Sheng Zhang(Principal Researcher, Microsoft Research), Hoifung Poon(General Manager, Microsoft Health Futures)