Agnieszka Szarecka wrote: > I am trying to run the loop structure refinement using loop_refine.py. > I have several 3-5 residue long gaps in my pdb structure and I am trying > to get them "patched". > > The program crashes with > > "..... ddihedr_262E> Central bond too short: 0.000"

This means that Modeller was unable to calculate the derivatives of a dihedral feature because the distance between the two central atoms in the dihedral became too short, leading to numerical instability. In this case, the distance is zero, meaning that the two atoms are on top of each other. This can happen if for some reason there's a strong attractive interaction between two atoms. You don't have electrostatics turned on, do you?

> Would anyone have an idea what the problem is and how I should fix it?

If you make no progress, tar up your input files and send them to me, and I'll take a look.

> Also: I ran the same time type of calcs with mod8v1 and had no such > problems. The only difference is that I switched from Linux to Windows > for this run.

The final structures you get are strongly dependent on the initial, randomized, coordinates. These differ between operating systems and Modeller versions due to differences in the representation and rounding of floating point numbers.

Ben Webb, Modeller Caretaker