But I still think the solution is to do the alignment internally - the user should be able to provide a PDB file and a target sequence, and the program will run MUSCLE automatically and figure out which row in the alignment output corresponds to each sequence. This avoids an extra step and would clear up the confusion regarding formats, ordering, etc.
-Nat
I agree this would be a solution to at least part of the problem. There is already some support for this, e.g. sequence { file_name = foo.seq chain_ids = A,B,C } would create an alignment internally using the sequence from foo.seq and chains A, B and C (and apply it to the same chains). It is not currently using muscle, but there is possibly no need - these are only pairwise sequence alignments. On the other hand, the alignment can be of crucial importance, and the user should not be limited to alignment algorithms implemented in Sculptor (and it is possibly not realistic to have an interface to all alignment programs, many of which are web services). Creating multiple sequence alignments via Sculptor (which may be more precise) would also complicate the user interface, for little apparent gain. BW, Gabor