I think there is a subtle issue with the foldchange.py example
In the api documentation the annotation volume is listed at as a 32-bit uint:
http://help.brain-map.org/display/mousebrain/API
But in the foldchange code, the data is loaded from the zip with the 'H' (unsigned short, 16-bit) designation.
I was running around in circles trying to figure out why the annot variable in the main() was twice as long as the gene-specific data. I believe this is the issue.
The solution is easy: change "arr = array.array('H', raw)" to arr = array.array('I', raw) in the DownloadAnnotationVolume() function
(reference: http://docs.python.org/2/library/array.html)