Sonidegib Phosphate

小分子的化学合成，一直是科学研究中的一项挑战。近年来，深度学习模型在逆合成领域展现出了独特的优势，这些技术能够很好的解决化学家在逆合成设计中的遇到的一些问题。然而，大多数基于AI的逆合成模型都缺乏可解释性，无法阐明逆合成设计中的内在逻辑。 2025年1月18日，西北工业大学生命学院施建宇教授团队在《Nature Communications》上发表文章Single-step retrosynthesis prediction via multitask graph representation learning。 这项研究介绍了一种革命性的方法--Retro-MTGR，该方法能够深度挖掘逆合成反应中的化学逻辑，让药物合成的逆向推理变得更加智能和精准。 逆合成推理 逆合成推理是从目标药物的分子结构出发，逆向推理出合成该药物的化学反应物。与传统的正向合成不同，逆合成推理让科学家们能够更高效地设计药物合成路线，极大地提高药物研发的效率。 Retro-MTGR：AI如何精准预测药物合成路径 施建宇教授团队提出的Retro-MTGR框架（如图1所示），是基于先进的“多任务图表示学习”的AI模型。通过深入学习分子结构中的图信息，Retro-MTGR框架能够同时完成多个任务，具体包括两个主要任务和一个辅助任务： 识别反应中心：判断分子中最容易发生化学反应的部分（主要任务）； 识别离去基：确定在反应中会“离开”分子的部分（主要任务）； 优化原子嵌入：为化学反应提供更加精准的原子表示（辅助任务）。 通过完成这三个任务，Retro-MTGR不仅帮助科学家们更好地理解化学反应机制，还能精准预测药物合成路径，为药物研发提供强大支持。 图1. Retro-MTGR的工作流程图 研究亮点：从键能到分子电性，AI揭示药物合成的深层规律 这项研究的最大亮点之一，在于通过学习分子化学键的结构，键能等属性，揭示了哪些化学键可能成为反应中心，哪些则可能只是普通的化学键（图2）。具体来说： 高键能（>=360 kJ/mol）的化学键，如双键和三键，通常是普通键； 低键能的单键（如C-N、C-O等）可能是反应中心，也可能是普通键，这取决于周围分子的结构； 反应中心的原子通常具有相反的电性，而普通键的原子则具有相似的电性。 不仅如此，Retro-MTGR还通过联合嵌入空间分析了合成子和离去基的电性分布，揭示了反应过程中的关键规律（如图3所示），这包括：反应中的两个合成子总是具有相反的电性并且彼此相距甚远；属于同一个反应物的合成子和离去基通常具有相反的电性并且彼此靠近；反应中同时出现的两个离去基经常具有相反的电性并且彼此相距甚远。 图2.化学键嵌入空间的展示 图3.合成子和离去基的联合嵌入空间展示 实验验证：准确预测药物合成路径 为了验证这一创新框架的有效性，研究团队通过预测两种药物Sonidegib和Acotiamide的逆合成路线，并与实际合成路线进行了对比。 图4. 对于两种药物的预测逆合成路线和真实化学合成路线展示 结果表明，Retro-MTGR推断出的合成路线与实际结果高度一致（如图4所示），这无疑证明了其在药物研发中的巨大潜力。 总结和展望 为了开发一个高度可解释的判别模型来揭示化学合成机制，本文提出了一个Retro-MTGR框架。基于分子图，在Retro-MTGR中同时考虑了三个相关任务，其中两个主要的监督判别任务负责识别反应中心和识别LGs，而一个辅助的自监督任务负责生成更好的原子嵌入。 与16种最先进的方法进行比较的结果，证明了Retro-MTGR的优越性。此外，在RTU和RTK场景中，不同的训练测试策略和不同大小的数据集验证了它的稳健性和可扩展性。涉及两种新药的实验验证结果表明，Retro-MTGR推断的逆合成路线与化学合成测定获得的逆合成路线一致。 Retro-MTGR是一个优秀的单步逆合成推断模型。未来，研究团队将在Retro-MTGR的基础上整合其他化学合成因子（例如反应产率，反应条件等）来开发一个可靠的多步逆合成模型以解决复杂分子逆合成路线规划的任务。 参考资料： Zhao, PC., Wei, XX., Wang, Q. et al. Single-step retrosynthesis prediction via multitask graph representation learning. Nat Commun 16, 814 (2025). https://doi-org.libproxy1.nus.edu.sg/10.1038/s41467-025-56062-y. --------- End --------- 感兴趣的读者，可以添加小邦微信加入读者实名讨论微信群。添加时请主动注明姓名-企业-职位/岗位或姓名-学校-职务/研究方向。

BURLINGAME, Calif.--( BUSINESS WIRE )-- Genesis Therapeutics , a company pioneering artificial intelligence (AI) technologies to create breakthrough medicines for patients, today announced the appointments of three key members during 2024 to its senior leadership team and Board of Directors: Shifeng Pan, Ph.D., as Chief Scientific Officer; Alla Ivanova as Senior Vice President of Engineering; and Paul A. Friedman, M.D., as Chairman of the Board. “ We are thrilled to welcome to our team these exceptional leaders who share belief in our mission and approach to transforming drug discovery with groundbreaking predictive and generative AI technologies,” said Evan Feinberg, Ph.D., founder and CEO of Genesis. “ Their industrial expertise across technology and drug discovery will enable us to further advance our GEMS (Genesis Exploration of Molecular Space) AI platform, expand our therapeutic pipeline, and ultimately create breakthrough treatments for patients suffering from severe diseases.” Shifeng Pan, Ph.D., Chief Scientific Officer: Dr. Pan joined Genesis in April 2024 and oversees R&D activities for Genesis’ pipeline of AI-enabled therapeutics. He brings more than two decades of experience in medicinal chemistry and biopharmaceutical research, including a 22-year career at Novartis, where he served as Head of Discovery Chemistry at the Genomics Institute of the Novartis Research Foundation. At Novartis, he led the discovery of FDA-approved drugs ODOMZO ® , BRAFTOVI ® , and MAYZENT ® , as well as numerous other programs across several therapeutic areas. Dr. Pan joins Genesis from Odyssey Therapeutics, where he was Executive Vice President and Head of Discovery. He earned a doctorate in organic chemistry from New York University. “ Genesis’ AI capabilities enable the discovery of novel and breakthrough molecules, making its platform well-suited to address challenging or historically undruggable targets,” said Dr. Pan. “ It’s an exciting time for the field of AI-enabled drug discovery and our premier team of engineers, chemists, and biologists who are leveraging the GEMS platform to rapidly advance our broad therapeutic pipeline.” Alla Ivanova, Senior Vice President of Engineering: Ms. Ivanova joined Genesis in November 2024 and brings over 25 years of experience driving innovation in software development; spanning data center management, large-scale infrastructure, and high-traffic web applications to cutting-edge AI/ML solutions. She has held executive leadership roles at notable tech companies including Head of Looker Engineering at Google Cloud, VP of Cloud & Data Infrastructure at Atlassian, Head of Core Infrastructure at Uber, Senior Director of Data Platform at Groupon, and Software Engineering Manager at Netflix. Ms. Ivanova earned her master's degree in applied mathematics and computer science from Kazan Federal University. “ At a time when AI is redefining the boundaries of what’s possible in drug discovery, I’m honored to be joining the Genesis team,” said Ms. Ivanova. “ The company’s AI platform, GEMS, has the potential to unlock the promise of computational drug design and discover life-changing medicines for patients. I’m excited to continue advancing transformational AI work, not just for innovation’s sake but to make a meaningful difference in human health – a mission that resonates deeply with me and the future I envision for AI.” Paul A. Friedman, M.D., Chairman of the Board of Directors: Dr. Friedman has served as Chairman of the company’s Board of Directors since April 2024. With decades of experience in biotechnology and drug development, he has held senior leadership roles at notable pharmaceutical companies, including CEO of Madrigal Pharmaceuticals (2016–2023), CEO of Incyte (2001–2014), President of Research and Development at DuPont-Merck Pharmaceutical, President of DuPont Pharmaceuticals Research Laboratories, and Senior Vice President of Merck Research Laboratories. He has also served as a board member for several publicly held pharmaceutical companies, including Incyte and Alexion. A diplomate of the American Board of Internal Medicine and member of the American Society of Clinical Investigation, Dr. Friedman earned his medical degree from Harvard Medical School. " Genesis is in an exciting period of growth following an outstanding Series B fundraise last year and recent collaboration with Gilead, positioning the company well to advance and expand its AI-enabled pipeline to develop breakthrough therapies for patients,” said Dr. Friedman. This expansion of Genesis’ senior leadership team follows the company’s appointments of two industry veterans to its Board of Directors: Vijay Pande, Ph.D., Founding General Partner at Andreessen Horowitz’s (a16z) Bio + Health fund, joined as a Director in 2023, and Mohamed “Sid” Siddeek, Vice President and Head of NVentures, NVIDIA’s venture capital arm, joined as an observer in 2024. About Genesis Therapeutics Genesis Therapeutics is pioneering AI for biotechnology by leveraging its generative and predictive AI platform called GEMS (Genesis Exploration of Molecular Space) for small molecule drug discovery. GEMS synthesizes proprietary AI methods, including language models, diffusion models, and physical ML simulations, to generate and optimize molecules for complex targets. Genesis has raised over $300 million from leading life science, tech, and AI-focused investors, and is building a therapeutic pipeline for a variety of high-impact targets. Genesis is headquartered in Burlingame, CA, with a fully integrated laboratory in San Diego. For more information on Genesis, please visit the company’s website at www.genesistherapeutics.ai .