Martini straight: Boosting performance using a shorter cutoff and GPUs

详细信息    查看全文

• 作者：Djurre H. de Jong^a ; Svetlana Baoukina^b ; Helgi I. Ingó ; lfsson^c ; Siewert J. Marrink^c ; ^{s.j.marrink@rug.nl" class="auth_mail" title="E-mail the corresponding author}
• 关键词：Gromacs ; Coarse-Grain ; Verlet neighbor search ; Lennard-Jones
• 刊名：Computer Physics Communications
• 出版年：2016
• 出版时间：February 2016
• 年：2016
• 卷：199
• 期：Complete
• 页码：1-7
• 全文大小：679 K

文摘

In molecular dynamics simulations, sufficient sampling is of key importance and a continuous challenge in the field. The coarse grain Martini force field has been widely used to enhance sampling. In its original implementation, this force field applied a shifted Lennard-Jones potential for the non-bonded van der Waals interactions, to avoid problems related to a relatively short cutoff. Here we investigate the use of a straight cutoff Lennard-Jones potential with potential modifiers. Together with a Verlet neighbor search algorithm, the modified potential allows the use of GPUs to accelerate the computations in Gromacs. We find that this alternative potential has little influence on most of the properties studied, including partitioning free energies, bulk liquid properties and bilayer properties. At the same time, energy conservation is kept within reasonable bounds. We conclude that the newly proposed straight cutoff approach is a viable alternative to the standard shifted potentials used in Martini, offering significant speedup even in the absence of GPUs.