From owner-chemistry@ccl.net Sun Jan 16 00:17:00 2011
From: "Andrew Dalke dalke*dalkescientific.com" <owner-chemistry*o*server.ccl.net>
To: CCL
Subject: CCL: PyMCS - still available?
Message-Id: <-43649-110115220440-30685-bwF7yAlEZ5fuJ3+N+q68iQ*o*server.ccl.net>
X-Original-From: Andrew Dalke <dalke===dalkescientific.com>
Content-Transfer-Encoding: 8bit
Content-Type: text/plain; charset=us-ascii
Date: Sun, 16 Jan 2011 04:04:23 +0100
Mime-Version: 1.0 (Apple Message framework v1082)


Sent to CCL by: Andrew Dalke [dalke_+_dalkescientific.com]
On Jan 13, 2011, at 10:44 AM, Kurt De Grave Kurt.DeGrave|*|cs.kuleuven.be wrote:
> The FOG program can compute maximal common subgraphs in polynomial time
> -- but only on outerplanar graphs. (It can tell you if one of the graphs
> is non-outerplanar, in which case you can fall back to another tool.)
> 
> http://people.cs.kuleuven.be/~jan.ramon/fog/

I looked into this because I've wondered about which types of search
can be improved by classifying structures into "easy to search" vs.
"hard to search", and using different algorithms as appropriate
rather than only using an algorithm which works for the hard case.

However, you don't make it easy for people to try out your software:

  - that URL does not have a download link, which is
      http://people.cs.kuleuven.be/~jan.ramon/fog/readme.html
    I found it by reading one of the PDFs

  - the readme says there is a tests/ subdirectory, but there isn't

  - the code uses "ushort" and "malloc.h" which aren't portable.
     I also had to force "CC=g++" in the Makefile in order to
     compile biconn.c (it's a C++ program with a ".c" extension)

  - While I have a copy of the NCI data set, and found the -db=
     option to change the default location for it, I didn't try
     to reverse engineer the format. It's neither SMILES nor SDF.

In what I can read it doesn't appear to handle what PyMCS handled,
which is the ability to generate the MCS from a group of compounds.

I know that other tools, such as GGA's free Indigo toolkit, have
code to solve this problem. My search for PyMCS is mostly emotional.
I wrote that code, and I spent about 6 weeks developing it. It was
the first non-trivial cheminformatics algorithm I wrote and it was
the basis for some of Bioreason's technology.

And sadly, no progress on tracking down that code.

Best regards,

				Andrew Dalke
				dalke!=!dalkescientific.com
-- 
Only a month left to register for my "Python for Cheminformatics"
and "Web Applications Development for Cheminformatics" courses.
14-15 and 16-18 February. Leipzig, Germany
http://dalkescientific.com/training/