Hi Pascal,
Nice to communicate with you here.
The standard base reference frame is described in the Olson
et al. (2001) JMB report. Specifically, the origin is defined as the center of what would a perfect Watson-Crick base-pair.
As detailed in the FAQ section of the current 3DNA homepage, the default set of parameters used in 3DNA was based on a survey of the NDB structures and is very
generous to account for distorted bps. As an example, how much a H-bond distance cut-off should be choosen? Intuitively, 3.2 A would be a more reasonable value than 4.0 A. However, even with such high cut-off, some users still report "find_pair" missing some bps, which turns out that the shorted possible H-bond distance would be over 5.0 A in such bps.
That's exactly the reason that 3DNA provides the "misc_3dna.par" file for users to tune it for their specific purpose. For example, by setting d_org = 1.0, dv = 0.5, z_ang = 15, one will find only Watson-Crick base-pairs.
As far as how stagger parameter is calculated, it is the projection of the vector linking the two base origins onto the mean base z-axis. A step-by-step description of the procedure can be found in the
1997 JMB SCHNAaP paper. 3DNA uses exactly the same algorithm except for a change in reference frame.
HTH,
Xiang-Jun