[Joelib-help] Re: [Joelib-devel] Descriptors and silent mode

[Joerg K. W. <we...@in...>]

Hi Nicolas,

> -> first, is there a silent mode, to avoid the printing of the descriptors
> loading ?
All debug, info, error messages can be switched off in the
joelib/src/log4j.properties
file.
So switch all DEBUG and INFO to ERROR and nothing will be printed.

> -> second, what is the meaning of the order of the values when for example
> you ask for "Atom_in_donor" descriptor and you have as output the following
> line: 5<0,0,1,0,0> ?
> Is this mean that the atom label 2 is donor ?
No, atom starts with 1,2,3, ..., N
So atom 3 is a Donor.
You can also use the general
joelib.molecule.types.AtomProperties
interface

> -> third, I have asked for Gasteiger-Marsili calculation: a first one is
> done without adding Hydrogens, and the second one with adding H in my
> molecules. But, at the end the results are different. What does it means ? I
> thought that this calculation take always the Hydrogens influence ?
Oh, this is a critical question and has no general answer.

 From the computer science standpoint of view you these are two cases 
and some descriptors are sensitive to hydrogens, others are not.
Why are some descriptors hydrogen independent ?
1. Some use only heavy weight atoms to calculate the descriptors.
2. Some simply remove the hydrogens, before they are calculated.

Of course, this is a serious problem, when having the case that one 
whants all sigma partial charges and on the other side you are 
interested in one special descriptor using only heavy weight atoms BUT 
here with the sigma partial charge atom property.
So this is and indexing conflict and of course can not be the same.

So, if calculating and accessing atom properties you can/should check 
the hydrogen atom state of the molecule, which is not big deal.
I've at least checked these things for the standard descriptor 
calculation class.

The easiest solution is to remove hydrogens, because they are still 
implicitely available via the JOELib expert system, but on the other 
hand the user might be interested in protonated molecules assigned by 
another program, e.g. Corina which is also really good to assign hydrogens.

> -> and fourth, how  could you explain that, for the same molecule:
> "Atom_in_donor" is 5<0,0,1,0,0>
> "Atom_in_acceptor" is 5<0,0,0,1,0>
> "Atom_in_donor_or_acceptor" is 5<0,0,0,0,0> ?
It's only one kind of possible and really huge discrete assignment/label 
space.
So, if you have a look at the used SMARTS pattern in and all literature 
references given there:
joelib.desc.types.AtomInDonAcc
joelib.desc.types.AtomInDonor
joelib.desc.types.AtomInAcceptor
see also
joelib.desc.types.HBA1
joelib.desc.types.HBA2
joelib.desc.types.HBD1
joelib.desc.types.HBD2
you can see that they use all different descriptions. And additionally 
then remind that SMARTS matching is based on the chemical expert systems 
used, so even if you use the same SMARTS, you will with a high 
probability get a nice number of mismatches.
E.g. i got once a time a complained from a NCI which compared the NCI 
and JOELib results. YES, that's a problem and i complain every time and 
in all publications, that we need here an 'chemical expert' system 
standard which we can test against.
http://www-ra.informatik.uni-tuebingen.de/software/joelib/tutorial/atomtyper.html

If you do not like them, create your own descriptor. If you want it more 
dynamic use SMARTS from a text file, but then i recommend to inroduce 
also an ID as done for the JOELib expert systems. This guarantees you 
that you will notice if another user changes on single SMARTS entry, 
which can definitely occur, because we are talking about expert systems. 
  Which are a really evolving type of label assignment (or more general 
model/hypothesis).

See here interfaces:
joelib.data.SoftCodedKernel
joelib.data.HardCodedKernel

Kind regards, Joerg
-- 
Dipl. Chem. Joerg K. Wegner
Center of Bioinformatics Tuebingen (ZBIT)
Department of Computer Architecture
Univ. Tuebingen, Sand 1, D-72076 Tuebingen, Germany
Phone: (+49/0) 7071 29 78970
Fax: (+49/0) 7071 29 5091
E-Mail: mailto:we...@in...
WWW:    http://www-ra.informatik.uni-tuebingen.de
--
Never mistake motion for action.
                                     (E. Hemingway)

Never mistake action for meaningful action.
                                (Hugo Kubinyi,2004)