Hi there,

1). There could be some (small) stochastic variation between FUBAR runs -- the methods in DM and HyPhy use the same code. Did you use the same posterior probability cutoffs (0.9 should be the default). Was the site that 'disappeared' in one of the runs close to 0.9 (e.g. 0.905?)

2). Datamonkey has some extra post-processing code to find sites under negative selection. You can get the list of such sites from FUBAR output by processing the .csv file (look at the Prob[alpha>beta] column)

3). Datamonkey reports the expected number of false positives, i.e. like FDR * [number of positive results]. What you choose to report is up to you. Generally I recommend reporting all the called sites, the cutoff value, and some FDR measure.

Sergei