Allosteric Regulation of Ectopically Expressed Olfactory Receptors in Tumor Cells: Ligand-Receptor Topology and Deep Learning-Assisted Drug Discovery

Abstract This research expands existing Olfactory Receptor (ORs) studies, exploring their potential in cancer drug therapy andestablishing a novel pipeline applicable to similar G protein-coupled receptors. Many ORs have been identified across ninecancer types and roughly 1000 cancer cell lines; their activation or inhibition of certain pathways can induce cancer celldeath and reduce proliferation/migration. Our initial focus was OR51E2, utilizing known ligands that modulate the receptor.Approximately 1.4M candidate molecules were sourced from the ZINC20 database based on molecular weight and refinedthrough Schrödinger Maestro’s multi-step virtual screening—including High-Throughput Virtual Screening, standard precision, and extra precision docking—to minimize Gibbs Free Energy upon binding. Compounds with the highest binding affinityunderwent descriptor analysis via Mordred, generating roughly 1,800 molecular descriptors each. Various classification modelswere then trained and tested on bioassay data from EMBL and DUD-E to predict ligand bioactivity. Predictions were finalizedthrough fine-tuning a Support Vector Machine and projecting molecular descriptors in 3D chemical space to improve accuracy.SwissADME assessed ADMET profiles for four promising candidates for OR51E2, each exhibiting drug-like characteristicsacross five metrics, with predicted bioactivity >85%. This study identifies ORs as promising cancer therapy targets, alongsideintroducing a novel drug discovery methodology.