Determine homology classes the same way we already do.
Then label the components with their homology class.
Then geometric hashing (on crosshairs only) to determine the correspondence,
based on when we get a hit, at a certain threshold.
Correspondence then allows RMSD to be determined, after extracting actual
coords.
Will prevent needing to extract C-alpha coords from all but the case that we
actually want to compute RMSD for. Should be much faster.
But this will only work when the model is correct. When it's incorrect, we
will not get a hit from the geometric hash, and couldn't do an RMSD then ...
Original issue reported on code.google.com by chad.a.davis@gmail.com on 31 May 2010 at 7:54
Original issue reported on code.google.com by
chad.a.davis@gmail.com
on 31 May 2010 at 7:54