Overview Topics

October, 2000

Variability in Current Simulations


Simulations are stochastic by nature. This means that there are variations between runs. Just as no two lab results are identical, no two simulations will be identical.

Blocking astrocytes greatly affect the overall neuronal health. By decreasing the diffusivity of chemicals locally, they slow down the spread of chemicals. This effect is felt most keenly on the concentrations of IL-6, a chemical that is secreted by astrocytes. Thus, IL-6 builds up greatly in the vicinity of blocking astrocytes. Because IL-6 is toxic to neurons, neuronal death becomes acute in these regions.

Astrocyte blocking is one of the most stochastic elements in the simulation. It depends on astrocytes being near fiber deposits. Because fiber deposits are placed at random, this greatly influences what astrocytes become blockers and when they may become blockers. In addition to this, the effects on the diffusivity are also stochastic. Thus, not all blocking astrocytes affect the spread of chemicals in the same way.

The following graph shows the variation inherent in the simulation:

Not a single parameter was changed between these runs. The final picture confirms that blocking astrocytes greatly influence the neuronal health.

Notice how the areas of greatest death (darker shades of gray) are near blocking astrocytes (blue discs near orange dots). Of course, the picture does not explain when the astrocyte became blockers which also adds to the variations.

When fibers are removed from the simulation, astrocytes cannot become blockers. In this case the variation between runs is much less as shown in the following graph:

By looking at the final picture, one can still see why some variation in the runs exist.

The variation is due in large part to the number of astroctyes that are close to the center. Because the wave of IL-1B usually spreads out from the center of the region where the microglia have gathered, astrocytes that are closer to the center get triggered first. Thus, for simulations where there is a greater concentration of astrocytes near the center, one would expect neuron death to happen a little more rapidly as more astrocytes release the toxin IL-6 sooner.

It is worth pointing out that the area of astrocyte effects is very new. Many of the rules and paramters we use are still being worked out. If the effects on diffusivity were much less than our current default, one would expect the variation between runs to decrease. Thus, the degree of variation may change as we change how astrocytes affect the system.

Knowing that variations exist, the auto-variant (or batch) simulation becomes a good way to test for these variations. Simply select the same minimum and maximum value for any parameter in the simulation. The result shows the variations between runs of identical parameters. Also, having this program written, makes future programs that analyze the variations in runs easier to produce.