Re: CCL: W:SMILES AND MODELLING

 Sent to CCL by: Geoff Hutchison [grh25(~)cornell.edu]
FIRST QUESTION: are SMILES the best choice for database entries and computation 
 of new structres???
 
 IMHO, no.
SMILES do not include any sort of 3D coordinates or thermodynamics information. SMILES also usually omits hydrogen atoms which may be quite important in your calculations.
Personally, I'd use CML (http://cml.sourceforge.net/) or even just an XYZ coordinate file where you store thermodynamics and other information in the 2nd "title" line. MDL SDF format (Mol file) might also be an excellent choice, since it's widely used and allows storing various calculated properties.
SECOND QUESTION: if you model spontaneous cyclization from a linear compound, how can you determine which product out of all possible ones is most likely to be the real thing
Well, since it sounds like you have the same atoms for all candidates you're examining, I'd look first at the computed total energy, (e.g.,
from a DFT B3LYP calculation using a decent basis set, or possibly a
semiempirical method like AM1, PM3 or PM5, although these may be less reliable depending on your system).
THIRD QUESTION: is there a software that can help us solve "second question", and could you recomend us some(it should cooperate with SMILES "Daylight toolkit" if possible)
There's a lot of computational chemistry software out there. You might consider (in no particular order) Gaussian, Q-Chem, GAMESS-US, Jaguar, MOPAC, HyperChem, Spartan... 
 
 P.S. are there any Java libraries for this stuff???
Sure. You might consider using CDK or JOELib, which can both read/ write to SMILES databases and handle transformations to other file formats, etc. Open Babel isn't Java, but it also helps reading/ writing to a wide range of chemical file formats. All of these are part of the "Blue Obelisk" movement for open source chemistry software. 
 http://www.blueobelisk.org/
 Cheers,
 -Geoff