arXiv 定量生物学· Hale S. Hasdemir (University of Maryland Institute for Health Computing, Bethesda, Maryland, USA), Venkata Sai Sreyas Adury (Chemical Physics Program, University of Maryland, College Park, Maryland, USA, Institute for Physical Science and Technology, University of Maryland, College Park, Maryland, USA), Pratyush Tiwary (Department of Chemistry and Biochemistry, University of Maryland, College Park, Maryland, USA, University of Maryland Institute for Health Computing, Bethesda, Maryland, USA)·· 6 天前AI 评分22
ANTIOCH-VS:从序列到候选化合物的模块化开源集成虚拟筛选框架
ANTIOCH-VS: From Sequence to Hits with Modular, Open-Source Ensemble Virtual Screening
AI 导读
马里兰大学团队发布开源虚拟筛选框架 ANTIOCH-VS,可从蛋白序列直接生成候选结合物排名短名单,无需配体结合晶体结构或人工干预。在 JAK2 激酶和 ADRB2 GPCR 的 DUD-E 基准集上,AF2RAVE 构象集合给出的 top-20 短名单命中率分别为 20% 和 60%(随机约 1-2%),与配体结合晶体相当,JAK2 上优于单一 AF2 模型和 apo 晶体。
来源:arXiv 定量生物学 · arxiv.org