I'll just a couple things to Peter's recommendations.
Yes, the Demo would probably be key to look at for starters-- there are a number of scripts in a ~pipe-liney-type ordering for making networks of target (functional) ROIs, and then for performing different forms of tractography through them. There are comments in the scripts, as well.
And yes, 3dDWUncert gets run before '3dTrackID -mode PROB'-- in order to do the probabilistic Monte Carlo iterations, one has to calculate the uncertainty of DTI parameters first.
I think tracking itself should be done in native space; so if you have functional/GM ROI networks in standard space (perhaps from a group study), then you can transform each of those into each individual's native DWI space (I usually use 3dAllineate), and perform tractography through them; afterwards, you can use the *.grid file structural matrices to easily combine values across the group. If you want to, there is also the map_TrackID program that Peter mentioned-- that will only be for visualization purposes for the tracks; you could also map the (volume) masks of the WM connection regions to standard space. For quantitative comparisons with the values of the WM ROIs, though, the *.grid files might be the best place to start -> and there are also the 3dMVM tie-ins now for combining tractography outputs and subject data in statistical models (see the @Install_FATMVM_DEMO).
--pt