GRID interaction with receptor

Unlike the other grid-based techniques, molecular descriptors are not derived from —> molecular interaction fields but from the partition of molecules into different parts that are expected to have different types of interactions with receptor sites. 

In order to characterise the three-dimensional design constraints imposed by the receptor, a first class of methods attempts to locate target sites within the binding cavity that will yield the crucial favourable interactions with predefined probes. The probes can be individual atoms - site points - or small fragments e.g. functional groups like carbonyl or methyl, water, which search for hot spots in the active site, as it is illustrated in figure 1. Examples of these methods are GRID [32, 33], MCSS [34], HSITE [35] and LUDl [36]. 

Atoms in molecules are first classified into different pharmacophoric types by the GRID force field parametrization. Features considered are hydrophobicity, hydrogen-bond donor and acceptor capabilities, and charge. Then, all accessible geometries for all the combinations of three or four features are calculated and encoded in the final vector. Fingerprints can be calculated both for ligands and proteins. In small ligands, pharmacophores are defined by triplets or quartets of atoms, vhich have critical interactions with a receptor. 

---