Large compound collections, virtual screening

Molecules in real solvents can exist in one or more tautomeric forms. The use of different tautomers in calculations can lead to significant variation in the estimated log P values (Pig. 15.1). Accurate prediction of the dominant tautomer requires ah initio calculahons. Due to speed limitations such calculations are not feasible for virtual screening and prediction of large compound collections. Moreover, the interpretation of the results can also be difficult, for example, the lacton-lactim (Pig. 15. IB) is the stable form of maleic hydrazide in the gas phase but the difference between this and the dilacton form (Pig. 15.1C and D) disappears in solution... 

The hit rate within a set of molecules selected by a virtual screen is primarily determined by two parameters the unknown proportion of p hits that exist in the set of molecules scored and the false positive error rate (a) of the classifier used for virtual screening. To a large extent, the statistics of rare events (true hits within a large compound collection) leads to some initially counterintuitive results in the magnitude of a hit rate within a set of molecules selected by a model. 

Availability of Large Compound Collections for Virtual Screening... 

CLEVER is a computational tool designed to support the creation, manipulation, enumeration, and visualization of combinatorial libraries. The system also provides a summary of the diversity, coverage, and distribution of selected compound collections. When deployed in conjunction with large-scale virtual screening campaigns, CLEVER can offer insights into what chemical compounds to synthesize, and, more importantly, what not to synthesize. In this chapter, we describe how CLEVER is used and offer advice in interpreting the results. 

One may think of an iterative model for the preclinical discovery screening cycle. A large number of compounds are to be mined for compounds that are active for example, that bind to a particular target. The compounds may come from different sources such as vendor catalogues, corporate collections, or combinatorial chemistry projects. In fact, the compounds need only to exist in a virtual sense, because in silico predictions in the form of a model can be made in a virtual screen (Section 8) which can then be used to decide which compounds should be physically made and tested. A mapping from the structure space of compounds to the descriptor space or property space provides covariates or explanatory variables that can be used to build predictive models. These models can help in the selection process, where a subset of available molecules is chosen for the biological screen. The experimental results of the biological screen (actives and inactives, or numeric potency values) are then used to learn more about the structure-activity relationship (SAR) which leads to new models and a new selection of compounds as the cycle renews. 

Second, the validated NMR methodology is applied to a screening approach at larger scale. Virtual screening creates the kinase-biased compound collection. Simplified models of interactions observed at the validation set provide the basis for the filtering rules, which downsize large in-house compound collections. 

---