A Distributed Parallel Computing Model of Blood Oxygenation

S. A. Williams, G. E. Fagg, P. C. H. Mitchell and
K. P. Williams

Departments of Chemistry,
PO BOX 225, Whiteknights,
The University of Reading,
Reading, UK, RG6 6AY

email: {shirley.williams, g.e.fagg}@reading.ac.uk

Abstract

The uptake of oxygen by blood is expressed experimentally as percentage saturation of blood against oxygen partial pressure. We wished to model the oxygenation uptake at the molecular level as a discrete event computer simulation. It was neces- sary to experiment with the simulation to determine the number of events and their associated parameters. A simple graphical user interface was developed that served as a workbench for running such simulations. The user can speculate on a process, quickly obtain the initial approximations, compare these with measured results and refine the original speculations.

Each experiment involved setting up a Monte Carlo simulation, that was "speeded up" by using PVM to spread the work across a distributed network of workstations. While a simulation was running the graphical user interface could be used to alter parameters and start other simulations that could be executed simultaneously, using a refined version of the PVM spawning procedure to ensure work was always distrib-uted to the least loaded machines in the cluster.

The workbench and corresponding PVM programs were designed so that the number of parameters for a particular model could be dynamically set at run time and varied from one experiment to the next. The number of runs per simulation, the number of processors and a variety of other parameters could also be dynamically changed between experiments, from the graphical user interface.

We experimented with data for blood from various products including: stripped human hemoglobin and shrimp hemocyanin. It was found in each instance that there was a three- or four-tuple of numbers associated with each molecule. Each of the numbers was associated with a pressure at which oxygenation would occur. The first representing a range 0 to infinity, the others representing smaller ranges between 0 and 1. For each number a "power" factor indicated the spread of numbers and a mul-tiplier served to match the units measured.

Using PVM we have developed a workbench that allows interactive investigation of the oxygenation process, using the spare capacity available across a distributed network.