pmapper and other (basic!) queries

Hi,





I've got 6 "hit" compounds from a high-throughput screen, and I want to try to generate a/some pharmacophore model(s) based on these structures.


Unfortunately, I don't really understand what PMapper is doing. What can I do with the output PMAP strings? When attempting to read them using the supplied colors.ini in MView, all the H atoms are yellow, as are a couple of other atoms (one N in one molecule, one S in another), and everything else is grey. Also, generating individual PMAP strings per molecule isn't what I'm after. Is there any way to compare all 6 molecules to work out some sort of pharmacophore based on all (or just some) of them?





I'm also unclear on how to use the PMAP strings to search other structures. I've been unable as yet to get a database set up (due in part to permissions problems), so this is somewhat hypothetical at the minute.





Sorry if these queries seem strange or overly basic - I'm completely new to java and am having enormous trouble understanding how to get the programs running. I take it there is no GUI for JChem base? I've not been able to find any kind of menu or instructions for this as yet, so I've been limited to pmapper at the command line and MarvinView/Space/Sketch.





For clarification, I do not want to write any web-based applications, in fact a simple, single, compiled executable would be the ideal thing for me!!





Thanks in advance,





Isaac

Hi Isaac,





no GUI indeed....





A good starting point to our pharmacophore tools is this document: http://www.chemaxon.com/jchem/doc/user/PFp2D.html This introduces the concept of PMAP and takes you one step further to pharmacophore fingerprints.





If you intend to use PMAP-s directly to construct a pharmacophore model (it is a viable approach) then you need to design and implement your own system. ChemAxon's tools do not use PMAP directly to construct a pharmacophore model, instead, PMAPs are used to generate pharmacophore fingerprints (see the document referred above) and these fingerprint can be combined for instance by the widely used consensus scoring to generate a pharmacophore model (or hypothesis). You can read more about pharmacophore hypothesis in the Introduction of http://www.chemaxon.com/jchem/doc/user/ScreenMD.html.





To search structures similar to you model - again, you need a hypothesis created from fingerprints, ie. you cannot search directly by PMAP - you need to use the ScreenMD software, see link above.





Regarding coloring by pharmacophore type, you find examples here: http://www.chemaxon.com/jchem/doc/user/PMapper.html#examples


If this does not help and you still experience odd colors, please provide some sample structure along with the corresponding PMAP you obtained by PMapper.





Hope this helps, please get back to the forum in case you need futher assictance.





Regards,


Miklos

Thanks for the response. After reading, trial and error, and a bit of cheeky-installing-of-MySQL, I've got everything up and running.





I have 6 "hit" compounds from a real HTS, and have generated the 2D pharmacophore fingerprints for these. I don't currently have access to the entire compound library (although this can be arranged if necessary).


I've read all of the documents I can find here, but I'm not clear on how (in practice) to use the fingerprints of these hits to generate a single pharmacophore hypothesis without using that to screen a virtual library. Ideally, I'd generate a structure/a few structures which represent the hypothesis, but I'm not sure if this is a practical step in the process.





The second query: what's the best method to select a training set from a compound library? Is there a program in JChem that will do this automagically, or should I generate a selection of random numbers and set aside 5% of the library manually (and indeed what is the practical method others use for this)?





Thanks again!





Isaac





PS: GUI or no GUI, I'm most impressed with what this software does!