Feature tree descriptor

This virtual search space can be searched using a feature tree descriptor. 

Fig. 4.1 The conversion of a molecule into a Feature Tree descriptor. The major phases are summarized on the right. A cyclic system is divided into individual rings only if this can be done uniquely. Features stored at the Feature Tree nodes are shape or chemistry related. The chemical feature (interaction potential) is color coded as follows red, H-bond acceptor blue, H-bond donor green, hydrophobic.

This subset was further investigated using the slower modeling methods to try to identify potential actives, known as plausible hits. An example of a molecule selected from the results of a docking experiment is shown in Fig. 4.8. This molecule had a similarity score of 0.93 to an active and is shown docked with the typical kinase inhibitor binding pattern. Both the active and the plausible hit are not drug-like from a medicinal chemistry perspective, but this example demonstrates well how the Feature Tree descriptor captures similarity between two molecules. 

So far, we have used the Feature Tree descriptor in pairwise comparisons. Most of the practical applications of molecular similarity can be reduced to the problem of comparing a pair of molecules. In some applications, however, it is extremely in-... 

Several methods are suited for navigation within such huge chemical spaces [56-59]. An extension of the above-mentioned feature tree descriptor makes it possible to search large virtual combinatorial libraries without enumeration [38]. 

The concept of feature trees as molecular descriptors was introduced by Rarey and Dixon [12]. A similarity value for two molecules can be calculated, based on molecular profiles and a rough mapping. In this section only the basic concepts are described. More detailed information is available in Ref. [12]. 

Note that no three-dimensional information is used for generating the feature tree the descriptor is therefore conformation independent. 

Finally, the Feature Tree nodes are marked with labels describing the shape and chemical properties of the building block. In principle, every kind of descriptor can be used as a label provided that the descriptor is additive over the building blocks. In our Feature Tree implementation, we normally work with a shape... 

Once a Feature Tree can be created from a molecule, the question arises of how to compare two Feature Trees. Using Eq. (1), we are able to compare two individual Feature Tree nodes. Owing to the additivity of the features stored at a node, we can also compare two sets of Feature Tree nodes. This is done by adding the features over all nodes within a set and applying Eq. (1) again. Obviously, we can also compare two complete Feature Trees in this way we just add all features in the two trees and apply Eq. (1). We call such a comparison level-0, because no division of the tree into pieces has been performed. Level-0 comparisons closely resemble the way linear descriptors work. If we assume for a moment that all components of a linear descriptor are additive and can be computed for each building block individually (such as the volume descriptor), adding the feature values over all Feature Tree nodes will create the linear descriptor. 

Substructure searching is often used in drug design and needs no further clarification. Similarity searching is also a very well known technique described in more detail elsewhere [52], We usually use M ACCS keys, Unity fingerprints, CATS descriptors, and feature trees for similarity searching [53], Each technique has its own strengths and weaknesses, so we favor parallel application of two or three of them. 

The following section gives an overview of feature trees and 2D autocorrelation vectors, the two most important graph-based topological descriptors used for virtual screening. 

A compromise between the 2D structural and 3D pharmacophore descriptors are techniques which implement reduced graph methods, whereby pharmacophoric elements are encoded in the 2D structure, thus gaining the benefit of a description which is more biologically relevant, whilst not adding the complication and noise from 3D conformations. Examples are the CATS, Reduced Graph,and Feature Trees approaches. 

---