Molecular: New state of the art hardware technologies and computational techniques to boost molecular simulations

Recent hardware developments have led to an extensive availability of multi-core CPUs and dedicated accelerator processors such as graphical processing units (GPUs). Using these devices, a massive parallelism at appreciable low cost can be implemented, producing an acceleration in computational performance that can be very high for several applications. To fully profit from this hardware enhancement an evolution of programs and algorithms is often required, reflecting in the demand of handling and major modifications of source codes. Program implementing MD simulations have usually a complex structure and their maintenance and development are not easy tasks.

The context of the project is an MD code named MDynaMix. It is a general purpose molecular dynamics code for simulations of mixtures of either rigid or flexible molecules, interacting by a force field in a periodic cell. MDynaMix is a diffused MD tools, widely used since the early 90’s for the study of several inorganic and biological systems. The aim of this project is to perform a major evolution of MDynaMix code, allowing it be up to date respect the new hardware and algorithmic releases that are now coming out. Particularly, the goal is to enhance MDynaMix to be GPU runnable on the class of GPU graphics boards based upon CUDA programming tool environment.

This implemented version of the MDynaMix code will be used to study in a systematic way the interactions between DNA and its alkaline physiologically counter ion. To verify the reliability of the results of the simulations two strategies will be employed. One involves the calculations of the Nuclear Magnetic Resonance relaxation parameters for the ions, to be compared with those obtained experimentally. Another involve the comparison of the results obtained varying the FF parameters used for the ions, and the water model, to check which MD setting lead to better results.