hi you can perform a pharmacophore modeling study by using common features pharmacophore for retrieving the common structural features from a set of active molecules. For this purpose you can use Phase (schrodinger) or Discovery studio (accelrys)

It depends on what you mean by pharmacophore.  The older-traditional medicinal chemists description is "chemical structural features (topological) responsible for biological activity"; or the modern definition "the pattern of chemical features (abstract) in 3D space responsible for a biological activity".  If you are talking about the former, there is a well-know algorithm called "Maximum Common Substructure" (MCS) that you can look up.  This will identify the common substructures that are common to all the compounds in your training set.  If you are asking about the latter, then most pharmacophore perception programs (like Phase, Catalyst, etc) has what is called "common feature hypothesis" algorithms that will do the job.  For both of these, however, 200 compounds is too large a set to gain any meaningful insight.  I would suggest you run some clustering algorithm to breakdown your 200 compound training set into smaller but more similar groups of compounds and run your pharmacophore perception programs over these clusters separately.

If you are looking for open source tool kits (there are also plenty of excellent proprietary ones

you can get away with just using OpenBabel, other good tools include RDKit and the Chemical development kit (CDK). If you are not a computer person per se you might appreciate Knime which which packages a more limited version of them together with many other tools in a workflow GUI. Note however:

You will be well served by spending the time to understand what exactly the program is doing and how.

I would also note that important underlying concepts and a good selection of related keywords are included in Osman Guner's post above. (ie. He is giving you what you need to know why those chooses are being made).

 All the above seems good  advice. It isn’t quite clear what additional information you might have to help, though. At very least, for discovering common van der Waal’s and electrostatic surfaces, not necessarily for lowest energy conformers, of conformationally flexible ligands, I used to use methods for incorporating algorithms due to Isidore Rigoutsos, Andrea Califano, and colleagues, e.g. Isidore Rigoutsos, Daniel E. Platt, Andrea Califano, and David Silverman. Representation and matching of small flexible molecules in large databases of 3d molecular information. In Pattern Discovery in Biomolecular Data, pages 111–129. Oxford University Press, 1999.

 If you have protein sequences for target proteins, then one should certainly not neglect taking account of these, especially if there are some homologous structures of known 3D structure, but of course sequence alone can still often help in modeling pharmacophores.  See for example Das R, Gerstein M.,”A method using active-site sequence conservation to find functional shifts in protein families: application to the enzymes of central metabolism, leading to the identification of an anomalous isocitrate dehydrogenase in pathogens.”, Proteins. 2004 May 1;55(2):455-63.

I tend to write own tools or put a battery of different available techniques together, since new algorithms were the point of the research in our case, so I can’t really speak for the benefits of any one modelling package.

you can use ligandscout software which is very useful programe in ligand-based pharmacophore and very simple to use