| Abstract
| - Conventional similarity searching of molecules compares single (or multiple) active query structures toeach other in a relative framework, by means of a structural descriptor and a similarity measure. While thisoften works well, depending on the target, we show here that retrieval rates can be improved considerablyby incorporating an external framework describing ligand bioactivity space for comparisons (“Bayes affinityfingerprints”). Structures are described by Bayes scores for a ligand panel comprising about 1000 activityclasses extracted from the WOMBAT database. The comparison of structures is performed via the Pearsoncorrelation coefficient of activity classes, that is, the order in which two structures are similar to the panelactivity classes. Compound retrieval on a recently published data set could be improved by as much as 24%relative (9% absolute). Knowledge about the shape of the “bioactive chemical universe” is thus beneficialto identifying similar bioactivities. Principal component analysis was employed to further analyze activityspace with the objective to define orthogonal ligand bioactive chemical space, leading to nine major (roughlyorthogonal) activity axes. Employing only those nine activity classes, retrieval rates are still comparable tooriginal Bayes affinity fingerprints; thus, the concept of orthogonal bioactive ligand chemical space wasvalidated as being an information-rich but low-dimensional representation of bioactivity space. Correlationsbetween activity classes are a major determinant to gauge whether the desired multitarget activity of drugsis (on the basis of current knowledge) a feasible concept because it measures the extent to which activitiescan be optimized independently, or only by strongly influencing one another.
|
