Vadim Patsalo wrote: > I am trying to take a pseudo-symmetric heterodimeric (AB) structure, > and build homodimer (AA) and (BB) structures. The chains have ~30-50% > identity, nearly the same in length. > The way I am doing this is aligning the AA sequence with the AB > sequence derived from the PDB with alignment.align(); However, > Modeller keeps doing something weird at the ends of the second chain, > placing breaks after the first and before the last residue. Chain A is > 3 residues shorter than chain B. Example, with underscores inserted > instead of the actual sequence: > > template ___DLH/SVIEN___VNLED/../.. > model ___DLH/H-IAN___VDL-H/../.. > > I would expect to have the model be "-HIA" in the beginning and "DLH-" > at the end. This is the alignment that I get from Clustal, and the > alignment that makes sense to me. I've tried setting 'overhang=1', > using a number of different matrices from '$LIB', to no avail. With > other matrices (Gonnet), I get a _horrible_ alignment, like so: > > VNLED/../.. > VDLH/-../.. > > How could it insert a gap into a HETATM chain? Is there a better way > to do this, perhaps aligning only the chains that are changed in the > sequence (align A to B), instead of AA with AB?

I wouldn't recommend using align() for multi-chain alignments - it has no special handling of the chain breaks. I'd recommend instead doing separate alignments of each chain and then combining the results manually.

Non-standard residues are scored the same as glycine, so I'd not recommend trying to align HETATMs. It makes little sense anyway.

Ben Webb, Modeller Caretaker