微软开源逆合成模型RetroChimera，登Nature
原标题：Improving synthesis prediction of small molecules at scale with RetroChimera
AI 摘要
微软研究院在《Nature》发表逆合成预测模型 RetroChimera，它结合基于 Transformer 的 R-SMILES 2 与基于图神经网络的 NeuralLoc 两个互补模型，通过学习式集成策略对二者预测排序。盲测中化学家更偏好 RetroChimera 给出的反应预测，优于子模型、既有方法甚至文献记录的反应。该模型已在 GitHub 以 MIT 许可开源并提供权重，同时可通过 Microsoft Foundry 访问，旨在加速药物与新材料研发。
正文节选
At a glance - We report on the recent publication of our retrosynthesis model RetroChimera in the journal Nature (opens in new tab). - The paper describes the model’s architecture as well as extensive validation studies, including the model’s ability to recall rare reaction types, and successful zero-shot transfer and fine-tuning on proprietary datasets. - We open-source RetroChimera’s implementation and weights in the hope that it will enable researchers to accelerate development of new medicin