Frequent hitter

Roche, O., Schneider, P., Zuegge, J., Cuba, W., and Kansy, M. Development of a virtual screening method for identification of Frequent Hitters in compound libraries./. Med. Chem. 2002, 45, 137-142. 

Figure 15.7 shows an analysis of the selectivity of2094 recent hits in high-throughput screens at Roche. Of these, almost half were specific for the screen in which they were identified as a hit - there were no other screens where these compounds had been titrated in a dose-response assay. However, 8% of the compounds had been tested in dose-response assays for more than five projects. These compounds are frequent hitters, and the most commonly occurring substructures are shown in Figure 15.8. Several of these compounds either are electrophilic, or can give electrophilic species on oxidation. The nitrobenzoxadiazoles are known fluorescent compounds. The oxindoles...

Fig. 15.8 Commonly occurring substructures among frequent hitters.

The application of computational algorithms for compound filtering and clustering is routinely used to eliminate undesired structures on the basis of chemically reactive fimctionalities, predicted liabilities (e.g., frequent hitters, hERG, cyp450 and so forth), or druglike properties. Computational methods are also used to group compoimds on... 

However, these compounds and the fragments are not without their intrinsic problems and should not be used as is. Some examples of potentially problematic compounds include those with chemically reactive groups, dyes, and fluorescent compounds which interfere with assays, frequent hitters/promiscuous binders, and inorganic complexes (55). It is important, then, to a priori filter out such compounds or reagents which are practically useless from a drug discovery point of view. 

Based on this in-house dataset, an in-silico prediction model [27] (three-layered neural network, Ghose and Crippen [28,29] descriptors) was constructed to evaluate the frequent hitter potential before compound libraries are purchased or synthesized. This model was validated with a dataset of the above-mentioned promiscuous ligands published by McGovern et al. [26], in which 25 out of 31 compounds were correctly recognized. 

Whereas hard filters can be considered to be knowledge-driven, soft filters are the result of a data-driven approach. A quantitative structure-activity or structure-property relationship (QSAR/QSPR) is established to predict a property from a set of molecular descriptors. Examples are the above-mentioned in-silico prediction tools for frequent hitters [27] and drug-likeness [41,42] additional models for ADM E properties are described below. 

The development of predictive models for drug-likeness, frequent hitters, ADME processes, and toxicological endpoints has so far yielded a great deal of soft filters (see discussion above and the compilation of ADMET computational models by Yu and Adedoyin [66]), and the trend still continues to improve both accuracy and... 

Figure 12.3 Using the example of the frequent hitters1 prediction model of Roche et al. [27], the probability that a flagged molecule actually has the indicated liability is calculated for a dataset with an even distribution between frequent and nonfrequent hitters (top) and for one with only 5% frequent hitters (bottom). In both scenarios it is assumed that frequent hitters are correctly classified with 96% and that the false positive rate equals 4%.

Roche O, Schneider P, Zuegge J, Guba W, Kansy M, Alanine A, et al. Development of a virtual screening method for identification of frequent hitters in compound libraries. J Med Chem 2002 45 137 12. 

Use computational methods that have been developed to rapidly and automatically identify potential frequent hitters (Roche et al., 2002 Seidler et al., 2003 Feng et al., 2005). Predictive models show some potential for predicting aggregation-based promiscuity in large libraries. A caveat of computational models is that they remain too crude to capture the concentration dependence of aggregate formation. 

---