Different scores when running locally compared to web interface

[nickhir]

Using the example data, I noticed that I get different interaction scores when I run the FlyPhone_parallel_batch.R on my local machine compared to the web interface. I know that the p values might be different because they are calculated using permutations, but I thought that the interaction scores should stay the same. Multiple runs of FlyPhone_parallel_batch.R always produce the same scores.
Multiple runs using the web interface produce slightly different results.

Any idea what might be the reason for this?

Cheers!