MARABOU

The structure generation phase, SPIDER, produces skeletons that contain no information concerning atom type. The vertices of the skeletons are described by hybridisation states alone and connections between vertices as bond types. Atoms must be substituted onto these skeletons to (1) promote binding of the ligand to the receptor, (2) to stabilise certain bonding situations or conformations, (3) to confer certain physical properties such as transport properties and (4) to facilitate ease of synthesis. This section describes the program MARABOU [1] that addresses the problem of atom substitution in order to confer the appropriate character at binding sites, e.g. hydrogen bond donor, hydrogen bond acceptor, etc.

Properties are assigned to target sites prior to structure generation and are used within this phase to ensure the appropriate heteroatoms are substituted for binding. The following properties can be assigned to target sites; hydrogen bond donor, hydrogen bond acceptor, either hydrogen bond donor or acceptor, positive charge, negative charge or neutral. These properties are either manually assigned to target sites or identified automatically using the HIPPO program.

Overview

The atom substitution program produces molecules by substituting combinations of functional groups onto the molecular skeletons. The figure illustrates the use of expert system technology to perform atom substitution. This provides a very flexible approach to the development of the program since all information concerning the functional group substitution is contained within a separate knowledge base. This library can be readily updated without recourse to re-programming. Each functional group substitution entry contains two parts. The first part is a description of the appropriate site of the skeleton onto which the heteroatoms of the functional group can be substituted. A linear string notation to describe molecular substructures is used to define this region. This language is based on the PATRAN language [2] developed for the LHASA synthesis design program and is also quite similar to the SMILES [3] notation. The second part of the functional group substitution entry is a rule describing the necessary atom and bond substitutions.

Rule

The above functional group substitution rule describes a three atom skeleton substructure onto which the appropriate atom substitution is performed. 'X' is used to denote a vertex, as the atom type of the skeleton is not defined at this stage and '-' and '=' are used to represent single and double bonds respectively. The atom features [HS=2] and [HS=3] define the number of attached hydrogens (assuming the skeleton vertices are carbon). This ensures there are no further connections on atoms 1 and 3 other than those defined. Properties of a vertex within the pattern, i.e. HACCEPTOR and NEGATIVE, must match the hydrogen bonding interaction or electrostatic interaction properties assigned to the vertices of the skeleton.

To generate a series of molecules from a particular skeleton, a number of operations are performed. Initially the skeleton structure is analysed and information concerning the target site properties, the chemically significant rings, the hydrogens attached to each atom (assuming each atom in the skeleton is a carbon) and the aromatic atoms and bonds is established. This information is generated to ensure correct mapping against the properties assigned to atoms and bonds within the PATRAN statements. The next stage is a substructure search where an attempt is made to match each PATRAN substructure statement against the skeleton. This produces a list of identified substructures that are used to fire the individual rules, resulting in a list of atom and bond substitutions corresponding to each valid rule. All combinations of functional group substitutions are generated and applied to the skeleton to produce a set of molecules. Combinations of functional group substitutions whose atoms overlap are discarded.

Example

This example illustrates the atom substitution phase in SPROUT. A skeleton structure has been generated to satisfy the six target sites shown. The required properties at each target site are presented. The program identifies a number of functional groups that map onto the skeleton. The three structures produced are generated through different combinations of functionality.

This section has described the atom substitution phase within the SPROUT program for de novo molecular design. For a given target skeleton, a series of molecules are produced containing the appropriate functionality to enhance the binding characteristics of the molecule. The program uses a knowledge base of 46 functional group substitutions. This provides a flexible approach to the future development of the program. Work is currently underway to allow further substitution by heteroatoms onto structures to enhance the ease of synthesis.

SPROUT typically generates thousands of potential structures and hence methods of prioritising sets of structures are essential to its practical use. The program CAESA has been developed to rank large sets of structures according to an estimate of synthetic ease. The program is an expert system and incorporates knowledge concerning various aspects of the structural complexity of molecules. Additionally the program is linked to a database of starting materials and a fast method of automatically selecting precursors has been developed. This information allows the program to identify structurally complex molecules that are easy to synthesise because of the availability of suitable starting materials.

References

1. P. Mata, V.J. Gillet; A.P. Johnson, J. Lampreia, G.J. Myatt, S. Sike, A.L. Stebbings, J. Chem. Inf. Comput. Sci., In Press
2. G. Hopkinson, Computer-Assisted Organic Synthesis Design. Ph.D Thesis, (1985)
3. D. Weininger, J. Chem. Inf. Comput. Sci., 30 (1990) 237.

You may want to

• Go on to ALLigaTOR
• Go back to SPIDeR
• Return to the abstract of the paper