Topological CATS descriptor

A particular property of the topological CATS descriptor is its speed of calculation. Thereby, the program qualifies for applications that deal with very large numbers of compounds, e.g. virtual screening campaigns in early stages of the drug discovery process. 

In a subsequent study, we examined the influence of seven similarity indices on the enrichment of actives using the topological CATS descriptor and the 12 COBRA datasets [31]. In particular, we evaluated to what extent different similarity measures complement each other in terms of the retrieved active compounds. Retrospective screening experiments were carried out with seven similarity measures Manhattan distance, Euclidian distance, Tanimoto coefficient, Soergel distance, Dice coefficient, cosine coefficient, and spherical distance. Apart from the GPCR dataset, considerable enrichments were achieved. Enrichment factors for the same datasets but different similarity measures differed only slightly. For most of the datasets the Manhattan and the Soergel distance... 

Which influence do different scaling methods have on the performance of the topological CATS descriptor We addressed this question with a comparison of three different ways of scaling the correlation vector descriptor [38] ... 

As stated previously for the topological CATS descriptor [31], the influence of different similarity metrics on the overall enrichment is marginal. For the full... 

The derivation of the topological distance matrix from the molecular graph is followed by the assignment of PPPs to the nodes of the graph. The following list provides chemical definitions of the five PPP types that are implemented in the CATS descriptor. The upper-case letter in parentheses is the abbreviation of each PPP type. Additionally, a functional group description is paired with its corresponding SMARTS in square brackets ... 

Figure 13.4 Results of a CATS similarity search. Similarity between the query structure (Haloperidol, a D2 antagonist upper left) and database compounds was defined in terms of a topological pharmacophore descriptor. The top 10 most-similar molecules found are shown.

Figure 13.12 A SOM-based pharmacophore road map. Different sets of ligands were projected onto a SOM that was generated by using the complete COBRA library. Black areas indicate the characteristic distributions of the compounds. Crosses indicate empty neurons in the map, i.e., areas of pharmacophore space that are not populated by the respective compound class. All molecules were encoded by a topological pharmacophore descriptor (CATS) [4], Note that each map forms a torus.

CATS descriptors are very similar to the PPP pair descriptors, the main difference being the topological distance betvsreen any pair of pharmacophore point types used in place of the geometrical distance [Schneider, Neidhart et al., 1999 Fechner, Franke ef al., 2003]. Moreover, tvhereas PPP pair descriptors are bit strings, CATS descriptors are —> holographic vectors where each bin encodes the number of times a PPP pair occurs in the molecule. 

CATS descriptors defined above are better named CATS2D descriptors because they are based on topological distances. 

Schneider et al. [125] developed and applied CATS descriptors as topological pharmacophores for scaffold hopping. This approach led to the prediction of novel cardiac T-type Ca channel blocking agents using mibefradil as starting structure. 

Naerum et al. [133] used topological pharmacophores to search for novel glycogen synthase kinase-3 inhibitors at Novo Nordisk. Using virtual screening based on the CATS descriptor, a novel chemotype with activity against GSK-3 was identified. After parallel synthesis around the identified motifs, interesting inhibitors with GSK-3 activity <1 [xM were found (Figure 12.11a). 

The following discussion considers three PPP pair descriptors CATS (topological PPP pairs), CATS3D (spatial PPP pairs) and SURFCATS (surface PPP pairs). Figure 3.1 provides a graphical overview. 

Fig. 3.1 The CATS family of descriptors CATS, CATS3D and SURFCATS. The degree of abstraction from the atomic molecular structure is assumed to be SURFCATS > CATS3D > CATS. All descriptors are based on a PPP-type description of the underlying molecule. For each descriptor, pairs ofPPPs are transformed into a correlation vector. CATS is calculated from the topological...

---