Editing Cealign plugin (section)

=== Fit vs. optAlign ===
====Take Home messages====
* [[fit]] and [[optAlign]] perform nearly equally as well
* if you need an algorithm with an appropriate reference, use [[optAlign]] (references at bottom of page).
* [[fit]] is faster -- if you're aligning many structures, use it over [[optAlign]]

====Discussion====
[[optAlign]] is a function within the [[Cealign]] package that performs the optimal superposition of two objects of equal length.  [[optAlign]] follows the Kabsch algorithm which is a closed form, and provably optimal solution to the problem.  [[fit]] on the other hand uses the Jacobi rotations to iteratively arrive at the solution of optimal superposition.  The difference in error between [[optAilgn]] and [[fit]] seems to be a non-issue (see below) as they both arrive at equivalent solutions for the rotation matrix.  The two algorithms are undertake different approaches to orthogonally diagonalizing the correlation matrix.

PyMOL's [[fit]] is fast and works well.  If you have to use something with a known reference then check out the "optAlign" function from the qkabsch.py file that comes with this [[Cealign]] package.  If not, you can just use [[fit]] and avoid installing new software.  :-)

optAlign is slower than fit.  I just tested both on a sample NMR ensemble; and, while not an extensive validation of "fit" it shows that (1) fit is faster; and (2) fit gets the same exact RMSD as "optAlign" (when optAlign is told to use all atoms, not just CA).  To make optAlign use all atoms and not just the alpha-carbon backbones, comment out (that is, put a "#" at the start of) lines 183 and 184 in qkabsch.py, where it says "CUT HERE."

<source lang="python">
fetch 1nmr
split_states 1nmr
delete 1nmr

# compare fit and optAlign RMSDs
for x in cmd.get_names(): print cmd.fit("1nmr_0001", x)
for x in cmd.get_names(): optAlign(x, "1nmr_0001")
</source>
<source  lang="bash">
# results from fit
0.0
4.50344991684
5.33588504791
5.78613853455
7.25597000122
6.67145586014
3.25131297112
3.36766290665
6.74802017212
5.1579709053
5.96959495544
6.68093347549
4.13217163086
5.51539039612
6.24266338348
6.03838825226
5.01363992691
5.33336305618
6.87617444992
7.797062397

#results from optAlign
RMSD=0.000000
RMSD=4.503450
RMSD=5.335886
RMSD=5.786138
RMSD=7.255970
RMSD=6.671456
RMSD=3.251313
RMSD=3.367663
RMSD=6.748021
RMSD=5.157971
RMSD=5.969595
RMSD=6.680934
RMSD=4.132172
RMSD=5.515390
RMSD=6.242664
RMSD=6.038388
RMSD=5.013640
RMSD=5.333363
RMSD=6.876174
RMSD=7.797062
</source>