[gmx-developers] Implementation of the Hungarian Algorithm on gmx rms (former g_rms)

[João M. Damas]

Dear developers,

I'll start by saying that I'm not requesting anything: it's not my
intention and it's frowned upon :-)

My intention is rather to call your attention to an algorithm that deals
quite elegantly with symmetry issues while calculating RMSD. You can see
here the paper with the implementation (even though the algorithm seems to
have been developed long ago): dx.doi.org/10.1021/ci400534h

While the advantages are quite obvious for the field of molecular docking
and structure-based design, I think a general tool such as gmx rms
(formerly g_rms) can benefit greatly from its implementation (they even
discuss it in the paper).

Recently I had to solve a symmetry problem while analysing the results of
high-throughput MD simulations of a ligand - protein system. Luckily, my
ligand only has planar symmetry with an inversion centre (and it was
impossible at this time to implement the aforementioned algorithm), so I
constructed the "symmetry mates" by scripting and calculated the RMSD
"symmetry-corrected" matrix, which I "fed" to g_cluster. A simple standard
RMSD vs. symmetry-corrected RMSD shows that the standard RMSD is
overestimating the distance between a great amount of pairs of structures
(ligand poses) due to the symmetry issue (see peak around [1,0.5]):

Of course g_cluster will create non-sense clusters or wrongly attribute
structures to clusters (I'm sorry, I cannot share those results).

This stands just as a proof-of-concept for the importance of such
algorithm. I'm remembering a few others, such as branching (bio)molecules
with symmetrical branches, but I'm sure there are a lot more applications
from this community that could benefit from such an implementation (as an
option, not as a replacement, of course).

Thank you for taking the time to read this.

Best regards,
João

-- 
João M. Damas
PhD Student
Protein Modelling Group
ITQB-UNL, Oeiras, Portugal
Tel:+351-214469613
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://maillist.sys.kth.se/pipermail/gromacs.org_gmx-developers/attachments/20150312/0259dbe0/attachment.html>