Validating Property and Metabolite Predictions for Some Novel Antimalarial Compounds
Robert D. Clark1
1Simulations Plus, Inc.
Malaria remains a major global health concern, and the limited resources available to the non-profit organizations working on new treatments makes it an especially good testing ground for drug design methodologies. Several years ago, Simulations Plus, Inc., undertook a proof-of-principle study in which literature data on dihydroorotate dehydrogenase inhibitors from the malarial parasite Plasmodium falciparum were used in conjunction with phenotypic screening data [1] to identify a promising series of 3-aminopropylamino-4-quinalone antimalarial leads. New analogs were designed based on both predicted activity and predicted ADMET properties. Seven such designed analogs were synthesized and four were found to be more active than the best of the lead series upon which they were based [2]. Physical properties of the four most active were determined and were found to be in good agreement with in silico predictions. They were also characterized in vitro with respect to five major cytochrome P450 (CYP) isoforms in a recombinant assay format and by incubation with intact human liver microsomes (HLM) in the presence and absence of NADPH. The metabolites formed in HLMs were analyzed by HPLC/mass spectrometry.
Those metabolites undergo a characteristic fragmentation in the mass spectrometer between their constituent quinolone, distal amine and bridging diaminopropyl units that makes it possible to localize oxidation sites within each molecule. Moreover, differences in predicted hydrophobicity makes it possible in several cases to rank-order the HPLC peaks with respect to likely elution position in the respective chromatographic profiles. Full synthesis would be required to definitively identify and quantitate all observed metabolites, but the observed variation in physicochemical characteristics and fragmentation behavior of these molecules makes it possible to disambiguate many of the HPLC/MS profiles. Doing so provides qualitative and quantitative validation of the predicted in vitro patterns of metabolism.
The procedure described can be seen as a prototypical workflow for assessing the accuracy of metabolite predictions in general as well as a demonstration of the qualitative and quantitative agreement between the metabolites obtained for the in silico predictions in this particular case. It illustrates how the multiple classification and regression models employed play off of each other in a realistic scenario as well as how ambiguities and apparent contradictions can be resolved.
1.Gamo, F.J. et al. Thousands of chemical starting points for antimalarial lead identification. Nature 2010, 465, 305-310.
2.Clark et al. manuscript in preparation.
