Computational Analysis of Receptor-Odor Interactions and Prediction of Behavior-Modifying Chemical Space
Coding of information in the peripheral olfactory system and resulting olfactory dependent behavior is thought to depend primarily on two fundamental factors: the interaction of individual odors with different subsets of the odor receptor (Or) repertoire, and the mode of signaling that an individual receptor-odor interaction elicits, activation or inhibition. In order to better understand these processes, we design and implement a structure-based virtual screening approach that identifies common structural features that are highly correlated with odor activity for individual receptors. We then apply these features to rapidly screen for putative ligands in silico from a large untested odor space (>240,000 putative volatiles) for the majority of odor receptors in the Drosophila antenna, allowing for analysis of odor coding for the majority of receptors for the first time. Functional experiments support a high success rate (~71%) for the screen and we validate numerous new activators and inhibitors for the receptors. Following our initial application in Drosophila, we extend our approach to predict activating and inhibiting odors for a large number of important pest and disease vector species including 50 Anopheles gambiae Ors (65% validated accuracy), the CO2 receptors of multiple species (48% validated accuracy), 9 newly identified Citrus Psyllid ORNs, and a large number of functionally distinct mammalian receptors. We next extended our in silico screening approach to identify shared structural features important for a behavioral response for DEET-like repellents which the molecular target has not yet been identified, identify ~150 natural compounds as candidate repellents. We select 4 candidates, 3 approved as safe for human food use, and demonstrate that they are strong olfactory and gustatory repellents to both mosquitoes and Drosophila. As only a small region of odor space has been explored, there remains potential to uncover previously unidentified patterns of odor coding. Through a combination of in silico and electrophysiology screens, we identify odors with ultra-prolonged termination kinetics that are delayed for several minutes, resulting in a memory trace that affects subsequent odor detection. Finally, we successfully perform structure- based virtual screen, identifying potential inhibitors of an important Tuberculosis drug target EthR.