Compound Selection and Database Filtering

It is one thing to have millions of individnal small molecules at one s disposal, but quite another to select the compounds yon want from this growing pool of diversity The first question to be answered is, of conrse, do we need to select compounds at all , because the answer is not always yes, we do Very often a random selection will perform as well as any pre-selected set, particularly when assaying orphan targets or those for which little or no information is available regarding endogenous ligands or 3D stmcmres. 

Such selections can be based, for example, on sub-structure similarity/dissimilar-ity calculations, or pharmacophore analyses each approach will generally lead to very different selections, even within a single set of compounds. The detailed concept of generating diversity has already been covered in this book, but it is valuable to review the basic concepts. 

For each molecnlar fragment that is present in the molecnle, 1 bit is set within the linear table. For example, the fingerprint above can be written digitally as  

Comparison of these fingerprints to a reference componnd is nseful in searching for either similarity or dissimilarity. The reference componnd itself can be chosen at 

Equation 1 Calculation of Tanimoto coefficients for similarity and dissimilarity 

---