摘要
针对传统生物医药学实验在进行糖尿病药物设计时的周期长、计算资源分散等问题,提出了采用高性能计算云的方式,将计算作为一种服务提供给用户,为开展分子动力学模拟等计算密集型任务提供可靠的解决方案。首先,利用分子可视化软件(VMD)构建两种人类胰高血糖素样肽-1受体蛋白(GLP-1R)分子体系;其次,以阿里云云计算平台为依托,采用分子动力学模拟计算软件(NAMD)对两种分子进行模拟计算,通过结果数据中的一系列相关性能指标对当下主流的同构、异构架构云主机的扩展瓶颈进行分析比较;最后,可视化两种分子的运动过程。理论分析和模拟计算的结果表明,异构计算优势较为明显,有20余倍的加速;但由于CPU与GPU、GPU之间的通信开销和负载不均衡等原因,使其扩展受到限制。实验观察到了GLP-1R受体和配体相互结合时精细动态相互作用的运动过程。该模拟结果不仅对研究者采用高性能计算云开展大规模分子模拟有一定的借鉴,而且对设计具有靶向性的II型糖尿病药物也有参考价值。
To solve long cycle and distribution of resources on diabetes drug design in traditional biomedical experiments,a high performance computing cloud providing computing as a service to users was proposed as a kind of reliable solution for such compute-intensive tasks.Firstly,two different-sized Glucagon-Like Peptide-1 Receptors(GLP-1R)were constructed by Visual Molecular Dynamics(VMD).Secondly,on Alibaba cloud computing platform,Nanoscale Molecular Dynamics(NAMD)was adopted to perform a molecular dynamics simulation experiment for the two molecules mentioned above,and a series of performance indicators obtained from the simulating result data,were used for evaluating performance and analyzing the bottlenecks homogenous multi-core CPU and heterogeneous CPU-GPU architecture.Finally,the motion of both molecules was visualized.Theoretical analysis and simulation results show that the advantages of heterogeneous computing are more obvious,with more than 20 times of acceleration.However,due to communication overhead and load imbalance between CPU and GPU,the expansion was limited.Meanwhile,the fine dynamic interactional process between GLP-1R protein and its ligand was observed.The simulation results have referential meanings to conduct large-scale molecular simulations experiments for the researchers,and to design the targeting drug used for treating type-II diabetes.
作者
戈孜荣
张洋
白启峰
张锦途
陈文波
GE Zirong;ZHANG Yang;BAI Qifeng;ZHANG Jintu;CHEN Wenbo(School of Information Science&Engineering,Lanzhou University,Lanzhou Gansu 730000,China;Office of Cybersecurity and Information,Lanzhou University,Lanzhou Gansu 730000,China;School of Basic Medical Science,Lanzhou University,Lanzhou Gansu 730000,China;College of Chemistry and Chemical Engineering,Lanzhou University,Lanzhou Gansu 730000,China)
出处
《计算机应用》
CSCD
北大核心
2019年第S02期179-183,共5页
journal of Computer Applications
关键词
高性能计算云
分子动力学模拟
阿里云云计算平台
胰高血糖素样肽-1受体
II型糖尿病
high performance computing cloud
molecular dynamics simulation
Alibaba cloud computing platform
Glucagon-Like Peptide-1 Receptor(GLP-1R)
type-II diabetes
