7.7.3 Comparison of the Libraries
Table 9 gives the hit rates for all example libraries, as well as the numbers of neurons matched by the compounds of the library.
The number of recovered hits in the screening libraries varies, and depends on a fortunate selection. Fig 26 shows the coverage of the descriptor space by the three designed libraries. The chemical space defined by fingerprints shows gaps, whereas genes and autocorrelation descriptors define a space that is homogeneously filled by the 768 selected compounds. The average density on the maps is four compounds per compartment.
•• •• • •• • •• •• ••• ••••
•1 ••• •«> • •• ••••••••• • •
O empty cell cell populated with 1 compound
• cell with highest population on the map
Figure 26. Kohonen maps of the diverse screening libraries. Gray levels indicate the population of a cell. Light gray cells contain only one compound; black cells indicate the highest population for each map (a: 141, b: 12, c: 8). The chemical space on the maps is defined by a) fingerprints, b) substructure descriptors, and c) autocorrelation coefficients (from three-dimensional structure).
Second-generation targeted libraries are designed to explore the chemical space near the hits in more detail. This is achieved by selecting compounds from the cells containing hits. The hit rates in the new libraries represent the capabilities of the different descriptor sets.
In case of the fingerprints, the hit rate is only slightly higher compared with the screening library or a random library. This shows that the population of active compounds in the hit neurons is not significantly higher than in any other region of the chemical space defined by the fingerprints. The fingerprint descriptor set therefore describes the structures of the compounds in a way that does not correspond to biological activity.
In contrast, the library designed on the basis of substructure descriptors as well as the autocorrelation-based library show a significant increase of hit rates. These descriptor sets can be used for identification of pharmacophore patterns and are applicable for selection of active compounds from a virtual library.
figure 27. Hit rales of designed and random libraries, targeted libraries
Was this article helpful?