Abstract

Drug discovery programs launched by the Medicines for Malaria Venture and other product-development partnerships have culminated in the development of promising new antimalarial compounds such as the synthetic peroxide OZ439 (Charman et al., 2011) and the spiroindolone NITD 609 (Rottmann et al., 2010), which are currently undergoing clinical trials. In spite of these recent successes, it is pivotal to maintain early phase drug discovery to prevent the antimalarial drug development pipeline from draining. Due to the propensity of the parasite to become drug-resistant (Muller and Hyde, 2010; Sa et al., 2011), the need for new antimalarial chemotypes will persist until the human-pathogenic Plasmodium spp. are eventually eradicated. Rational post-genomic drug discovery is based on the screening of large chemical libraries – either virtually or in high-throughput format – against a given target enzyme of the parasite. Experimental tools to validate candidate drug targets are limited for the malaria parasites. Gene silencing by RNAi does not seem to be feasible (Baum et al., 2009). Gene replacement with selectable markers is (Triglia et al., 1998), but it is inherently problematic to call a gene essential from failing to knock it out. However, none of the reverse genetic methods is practicable at the genome-wide scale. On the other hand Mestres et al. (Cases et al., 2005; Mestres et al., 2006) have annotated a library of molecules targeting NHRs. Using a hierarchical classification for 200.000 ligands and 5 receptors, chemogenomic links bridging ligand to target space can be easily recovered to distinguish selective from promiscuous scaffolds. Using Shannon Entropy descriptors (SHED) based on the distribution of atom-centred feature pairs, any compound collection can be screened to identify hits presenting SHED distances to a reference NHR ligand beyond a defined threshold and therefore likely to share the same NHR profile. Here, we successfully applied a machine-learning algorithm using Bayesian statistics (Xia et al., 2004) to predict target profiles from extended connectivity conserved motif like binding site active pharmacophore fingerprints of selected compounds from the biologically annotated free and non commercial databases (Nidhi et al., 2006) in resulting finally to a potent computer predicted inhibitor comprising potential hyper-mimicking activities to 5 conserved anti-plasmodium peptides.