Molecule-pharmacophore superpositioning and pattern matching in computational drug design Part 2

The term ‘3D pharmacophore’ 

The term ‘pharmacophore’ has become increasingly used in medicinal chemistry in recent years and has had different meanings attributed to it. ‘Pharmacophores’ are often regarded as structural fragments or functional groups being related to a chemical compound. However, the official IUPAC definition from 1998 is more precise: ‘A pharmacophore is the ensemble of steric and electronic features that is necessary to ensure the optimal supramolecular interactions with a specific biological target structure and to trigger (or to block) its biological response’. This definition clearly emphasizes the abstraction of common steric and electronic interactions of bio-active compounds exhibiting comparable biological effects within the same binding site in a comparable situation. This abstract model, containing chemical functionalities (such as ‘positive ionizable’ instead of ‘primary amine’) can serve as an effective search filter for virtual screening.This concept is not new in medicinal chemistry and has already been successfully applied before computers were used in chemistry. 

Hydrogen bonding interactions

                            Hydrogen bonding occurs when covalently bound hydrogen atoms with a positive partial charge interact with other atoms with a negative partial charge. To capture the characteristics of hydrogen bonding, Catalyst and LigandScout model H-bond donor and acceptor features as a position for the heavy atom and a projected point representing the position from which the participating hydrogen will extend. These two positions form a vector that indicates the direction from the heavy atom to the projected point of the hydrogen bond. However, in Catalyst only a single hydrogen bonding feature is permitted per heavy atom, whereas LigandScout allows an acceptor or donor atom to be involved in more than one H-bonding interaction.  

Lipophilic areas

                         Lipophilic contacts represent layer 4 features with no geometric constraints and are generally represented as tolerance spheres located in the center of hydrophobic atom chains, branches or groups. Although the perception of hydrophobic areas in Catalyst, Phase and LigandScout is based on the same algorithm described during Catalyst development by Green et al., subtle deviations seem to exist, and the results differ considerably, which makes an otherwise possible program interoperability hard to accomplish. In contrast to LigandScout and Catalyst, which place a hydrophobic feature shifted away from the ring center toward lipophilic atoms, Phase and MOE do not recognize a hetero aromatic ring as a hydrophobic area. For hydrocarbon chains, the number of hydrophobic features recognized by Catalyst, LigandScout and Phase depends on the length of the chain. MOE generally generates three annotation points per chain with one point in the middle and one on each end. This rough representation of lipophilic information in MOE makes it difficult to correctly describe lipophilic features for virtual screening – the dynamic placement as described by Green et al. allows for a smoother overlay of lipophilic areas, such as aromatic rings, isopropyl moieties, and aliphatic chains. In MOE, currently there is no such possibility using the built-in feature definitions, but there is a possibility to define one’s own algorithms, using the scripting language, SVL, provided with the software package. Scripting chemical feature placement, however, negatively influences virtual screening performance.