If I understand what you're attempting to accomplish, as David said, this is a vigorous area of study in structural biology. I'll start by saying that a single amino acid substitution will often make virtually no difference in a structure, but there are proteins where a couple substitutions will convert an all alpha-helical protein to all beta-sheet, so I'll leave it to you to judge what's worth trying. There are three routes you can take:
First, for small changes, pymol does have a built in force field ("molecular sculpting") and residue mutagenisis. Chimera has a tool called "minimize structure" that are probably better developed. And VMD has the force field tool kit plugin.
If you expect larger changes, you might go with a full simulation package like NAMD or CHARMM. They would take significant effort to get running, but they're there.
Finally, you can just go with an empirical structure prediction server. All you do is open your sequence file, make your substituions there, upload it to the server, and wait for them to email you the results. They'll do sequence alignments to find relevant PDB files and everything. I recommend the Robetta server (running David Baker's Rosetta). http://robetta.bakerlab.org/