Use of Machine Learning to Identify Novel, Behaviorally Active Antagonists of the Insect Odorant Receptor Co-Receptor (orco) Subunit.

Olfaction is a key component of the multimodal approach used by mosquitoes to target and feed on humans, spreading various diseases. Current repellents have drawbacks, necessitating development of more effective agents. In addition to variable odorant specificity subunits, all insect odorant receptors (ORs) contain a conserved odorant receptor co-receptor (Orco) subunit which is an attractive target for repellent development. Orco directed antagonists allosterically inhibit odorant activation of ORs and we previously showed that an airborne Orco antagonist could inhibit insect olfactory behavior. Here, we identify novel, volatile Orco antagonists. We functionally screened 83 structurally diverse compounds against Orco from Anopheles gambiae. Results were used for training machine learning models to rank probable activity of a library of 1280 odorant molecules. Functional testing of a representative subset of predicted active compounds revealed enrichment for Orco antagonists, many structurally distinct from previously known Orco antagonists. Novel Orco antagonist 2-tert-butyl-6-methylphenol (BMP) inhibited odorant responses in electroantennogram and single sensillum recordings in adult Drosophila melanogaster and inhibited OR-mediated olfactory behavior in D. melanogaster larvae. Structure-activity analysis of BMP analogs identified compounds with improved potency. Our results provide a new approach to the discovery of behaviorally active Orco antagonists for eventual use as insect repellents/confusants.