Parallel Algorithms for the Simulation of Protobiological Membranes

Objective: Our objective is to develop efficient algorithms for molecular-level computer simulations of membrane systems. Ultimately, we want to simulate the structure and functions of bilayer vesicle systems, which were precursors of modern cells.

Approach: We use molecular dynamics computer simulations to study water and water-membrane interfaces. Since interfacial systems are anisotropic, load-balancing using a parallel link-cells algorithm for short-ranged forces is applied to calculate efficiently the forces which act between all the molecules in the system. Long-ranged effects are included by employing multipole expansion methods.

Accomplishments: Pure water ice interfaces--To test the performance of our code for anisotropic systems, we have carried out computer simulations of bulk amorphous ices, and the amorphous ice formed from vapor deposition onto a cold substrate. For this homogeneous system, we have found that about 7 nodes on the IBM SP-2 are needed to achieve single-processor CRAY C90 performance and the force calculation is scalable to 64 nodes on the SP-2. This work has been carried out as part of a study of amorphous ice and is in press: "High density amorphous ice, the frost on interstellar grains." P. Jenniskens, D. F. Blake, M. A. Wilson, and A. Pohorille, Astrophysical Journal, in press, 1995.

Water-membrane systems--We have carried out computer simulations of water-membrane interfaces and the transport of small molecules across these interfaces. For the full water-membrane system, we must use about 8 nodes on the IBM SP-2 to achieve the single-processor performance we get on a CRAY C90. Currently, the scalability is good up to only 32 processors.

Significance: The study of amorphous ice yielded information about the structure of its "high-density" and "low-density" forms, which are the most common states of water in the universe. In particular, it was shown that the high-density ice can be described as a collapsed lattice of the more familiar low-density ice. In our simulations of the transport of small molecules across water-membrane systems it was demonstrated that many of these molecules exhibit increased concentrations, adopt specific orientations, and ultimately participate in chemical reactions with increased probabilities. These properties are significant not only for protobiological evolution but also for drug delivery and the mechanism of anesthetic action.

Status/Plans: Although we were successful in applying scalable algorithms to simulate membrane fragments, it is not clear whether we will be able to perform similar calculations for a full membrane vesicle system. This objective requires algorithms that are efficient and scalable to a LARGE number of processors. We intend to address this issue of scalability by exploring several novel algorithms, particularly particle-particle particle-mesh (PPPM) methods. Also, we plan to port the code to the CRAY T3D computer. We will investigate primarily the structure and cellular functioning of proteins incorporated into membranes.

Figure Caption: The figure shows the porous structure that results when water molecules at 77 K impinge on an amorphous ice substrate at 77 K. While there are many voids in the structure, the water molecules are, for the most part, tetrahedrally ordered. Most of the voids disappear when the structure is annealed.

Points of Contact:

Andrew Pohorille
Michael A. Wilson
NASA/Ames Research Center
pohorill@raphael.arc.nasa.gov, 415-604-5759
mwilson@max.arc.nasa.gov, 415-604-5496