准确而又快速的蛋白质分子动力学模拟是生命科学、生物医药和生物能源等领域中最具挑战和前沿的研究课题之一。描述原子间相互作用物理模型的准确性和蛋白质构象取样的效率是上述计算成功的决定性因素。目前蛋白质分子动力学模拟采用经典分子力学模型因而很难精确描述细致微观相互作用；同时，现有蛋白质分子动力学模拟并行计算由于构象取样效率低而限制了计算结果的收敛速度。我们前期改进了副本互换蛋白质构象增强型采样方法并与量子力学精度蛋白质能量分块计算方法结合，在天河II机器上完成了28万CPU主核测试，并行效率达到32%。本项目将立足于上述研究基础，针对天河II和国产Exascale众核机器，提出更高效的构象采样方法，结合量子化学计算程序的众核移植和优化及分块算法的改进，实现量子力学精度蛋白质分子动力学模拟的大规模并行算法和软件。为我国超级计算机研发奠定坚实的理论基础以及上述领域的应用研究提供更好的研究工具。

Accurate and fast molecular dynamics (MD) simulation of protein is one of the most challenging and cutting-edge technologies in life sciences, biomedicine and bioenergy research field. The success of the MD simulation critically relies on how accurate the interaction between atoms can be described by a physical model as well as how efficient the sampling of protein conformations can be achieved. Current MD simulation of protein is based on classical molecular mechanics (MM) model, resulting in inaccurate description of the inter-atomic interactions on microscopic level, meanwhile, the convergence of parallel computing is largely limited due to low efficiency of sampling techniques. In our previous studies, we have improved the traditional replica exchange method, which was further combined with the quantum mechanics (QM) calculation of molecular fragments of protein. This hybrid technique has been initially tested with 280 thousand cores on the Tianhe II supercomputer, achieving an efficiency of 32%. Based on our research experience and preliminary results mentioned at above, we propose to develop more efficient conformational sampling technique associated with computers, such as TianHe II and Exascale supercomputers assembled by made-in-china CPUs, through the migration and optimization of quantum chemistry program to the multi-core architectures as well as the improvement of molecular fragment calculation algorithm, such that we are able to accomplish the task of developing the large scale parallel algorithm and software of MD simulation of protein with QM-level accuracy. The improved parallel algorithm and software will lay a solid theoretical foundation for the development of supercomputers in our country, and at the same time provide better research tools for many related research areas.