Understanding the behavior of biomolecules such as proteins requires understanding the critical influence of the surrounding fluid (solvent) environment--water with mobile salt ions such as sodium. Unfortunately, for many studies, fully atomistic simulations of biomolecules, surrounded by thousands of water molecules and ions are too computationally slow. Continuum solvent models based on macroscopic dielectric theory (e.g. the Poisson equation) are popular alternatives, but their simplicity fails to capture well-known phenomena of functional significance. For example, standard theories predict that electrostatic response is symmetric with respect to the sign of an atomic charge, even though response is in fact strongly asymmetric if the charge is near the biomolecule surface. In this work, we present an asymmetric continuum theory that captures the essential physical mechanism--the finite size of solvent atoms--using a nonlinear boundary condition (NLBC) at the dielectric interface between the biomolecule and solvent. Numerical calculations using boundary-integral methods demonstrate that the new NLBC model reproduces a wide range of results computed by more realistic, and expensive, all-atom molecular-dynamics (MD) simulations in explicit water. We discuss model extensions such as modeling dilute-electrolyte solvents with Debye-Huckel theory (the linearized Poisson-Boltzmann equation) and opportunities for the electromagnetics community to contribute to research in this important area of molecular nanoscience and engineering.
Performance of universal machine-learned potentials with explicit long-range interactions in biomolecular simulations
NEC Laboratories Europe GmbH（NEC欧洲实验室）；NEC OncoImmunity AS（NEC OncoImmunity公司）；Institute for Artificial Intelligence, University of Stuttgart（人工智能研究所，斯图加特大学）
Long-range interactions & parallel scalability in molecular simulations
典型的生物分子系统如细胞膜、DNA和蛋白质复合物是高度带电的。因此，有效和准确地处理静电相互作用是非常重要的计算建模这样的系统。我们采用GROMACS仿真软件包，在一系列计算机体系结构上对不同的常用静电方案进行了广泛的基准测试（Pentium-4、IBM Power 4和Apple/IBM G5），单处理器和并行性能最高可达8个节点-我们还测试了四种不同网络上的可扩展性，即Infiniband、GigaBit以太网、快速以太网和几乎统一的内存架构，即，通过直接从其它CPU的本地存储器读取或写入其它CPU的本地存储器，CPU之间的通信成为可能。事实证明，粒子网格埃瓦尔德方法（PME）的表现令人惊讶的好，并提供有竞争力的性能，除非并行运行的PC硬件与旧的网络基础设施是必要的。使用大小为128、512和2048个脂质分子的脂质双层作为代表生物分子模拟中遇到的典型情况的测试系统。我们的研究结果使一个准确的预测计算速度在大多数当前的计算系统，无论是串行和并行运行。这些结果应有助于例如为一个小型部门计算机集群选择最合适的配置。
