The paper describes the generation of four types of three-dimensional molecular field descriptors or 'field points' as extrema of electrostatic, steric, and hydrophobic fields. These field points are used to define the properties necessary for a molecule to bind in a characteristic way into a specified active site. The hypothesis is that compounds showing a similar field point pattern are likely to bind at the same target site regardless of structure. The methodology to test this idea is illustrated using HIV NNRTI and thrombin ligands and validated across seven other targets. From the in silico comparisons of field point overlays, the experimentally observed binding poses of these ligands in their respective sites can be reproduced from pairwise comparisons.

Virtual screening is being applied to reduce the high-throughput screening bottleneck in many pharmaceutical companies and to reduce compound wastage. Cresset's ligand-based virtual screening technology using molecular fields can facilitate rapid identification of novel chemotypes from biologically testing only 200 - 1000 compounds. Four molecular fields calculated using the interaction of different probe atoms with the ligand are sufficient to describe how a ligand binds to its protein. Compounds with similar fields to known active ligands are predicted to have a high probability of showing similar activity. As binding is related to field similarity, this property has been exploited further to predict the bioactive conformation of small sets of structurally diverse active ligands starting from the two-dimensional structures alone without knowledge of the target site structure.

We have previously described the design and development of CI-988, a peptoid analogue of CCK-4 with excellent binding affinity and selectivity for the CCK-B receptor. Due to its anxiolytic profile in animal models of anxiety, this compound was developed as a clinical candidate. However, during its development, it was determined that CI-988 had low bioavailability in both rodent and nonrodent species. In the clinic, it was further established that CI-988 had poor bioavailability. Thus, there was a need to identify an analogue with an improved pharmacokinetic (PK) profile. The poor bioavailability was attributed to poor absorption and efficient hepatic extraction. We envisaged that reducing the molecular weight of the parent compound (5, MW = 614) would lead to better absorption. Thus, we synthesized a series of analogues in which the key alpha-methyltryptophan and adamantyloxycarbonyl moieties, required for receptor binding, were kept intact and the C-terminus was extensively modified. This SAR study led to the identification of tricyclo[3.3.1.1(3,7)]dec-2-yl [1S-[1 alpha(S*)2 beta]-[2-[(2-hydroxycyclohexyl)amino]-1-(1H-indol-3- ylmethyl)-1-methyl-2-oxoethyl]carbamate (CI-1015, 31) with binding affinities of 3.0 and 2900 nM for the CCK-B and CCK-A receptors, respectively. The compound showed CCK-B antagonist profile in the rat ventromedial hypothalamus assay with a Ke of 34 nM. It also showed an anxiolytic like profile orally in a standard anxiety paradigm (X-maze) with a minimum effective dose (MED) of 0.1 microgram/kg. Although the compound is less water soluble than CI-988, oral bioavailability in rat was improved nearly 10 times relative to CI-988 when dosed in HP beta CD. The blood-brain permeability of CI-1015 (31) was also enhanced relative to CI-988 (5). On the basis of the overall improved pharmacokinetic profile as well as enhanced brain penetration, CI-1015 (31) was chosen as a development candidate.

Phospholipase D (PLD) activity has been shown to be GTP-dependent both in vivo and in vitro. One protein that confers GTP sensitivity to PLD activity in vitro is the low-molecular-mass G-protein ADP-ribosylation factor (Arf). However, members of the Rho family and protein kinase C (PKC) have also been reported to activate PLD in various cell systems. We have characterized the stimulation of PLD in HL60 cell membranes by these proteins. The results demonstrate that a considerable proportion of HL60 PLD activity is located in a detergent-insoluble fraction of the cell membrane that is unlikely to be a caveolae-like domain, but is probably cytoskeletal. This PLD activity required the presence of Arf1, a Rho-family member and PKC for efficient catalysis of the lipid substrate, suggesting that the activity represents PLD1. We show that recombinant human PLD1b is regulated in a similar manner to HL60-membrane PLD, and that PKCalpha and PKCdelta are equally effective PLD activators. Therefore maximum PLD activity requires Arf, a Rho-family member and PKC, emphasizing the high degree of regulation of this enzyme.