On 5/28/10 4:24 AM, Claudio M. Soares wrote: > I am trying to model a really large structure with a lot of chains > (180 to be more exact - this is a complete virus capside) and I am > encountering problems with the PDB format used in the automodel class. > I manage to convince it to read a PDB file with 4 false chains > (corresponding each to 45 "real" chains). However, the script finds > out 180 chains internally (so much for fooling it...), writes them to > the *.INI file with bogus chain IDs (and bogus residue and atom > numbers) that it cannot read again when trying to produce the models. > How to solve this? Can I change the automodel class to read mmCIF > instead of PDB?

Unfortunately Modeller uses PDB files more or less exclusively, so you are out of luck for now. (It can write mmCIF files, but not read them, and the internal limits on residue numbers match those in the PDB format.)

An alternative might be to model just one subunit of the capsid (perhaps with enough neighbors to ensure it gets the right subunit-subunit contacts) and assemble the model subunits into the capsid later manually, e.g. with the Chimera package.

Ben Webb, Modeller Caretaker