In the human genome, proteins encoded from ~25,000 genes are designated as the prime drivers of complex biological processes. Considering a cell's molecular, functional, and physiological complexity, the required number of functional proteins outnumber the ones encoded by the genome. Moreover, this complexity is further exaggerated in disease states such as cancer, where the overall functional networks are much more sophisticated and, therefore, to ensure survival, cancer cells selectively activate a set of ectopic genes whose functional importance, at least in the context of cancer biology is very limited. One such example of a gene family is that of Odorant Receptors (ORs). In addition to their expression in the sensory epithelium of the nose, almost 92.5% (370/400) of the ORs are reported to be expressed in non-olfactory tissues, both in homeostatic and pathological states. We recently traced the expression of ORs at the single-cell resolution across multiple cancer types and observed their potential relevance with cancer cell differentiation status and prognosis. To identify their potential (non)agonists, we created one of the largest datasets encompassing GPCRs-ligand information across vertebrates. We utilized this gigantic database to develop state-of-the-art Machine Learning architecture to deorphanize orphan GPCRs (including ORs). Our Model (EvOlf) identified putative ligands for orphan GPCRs, of which a subset of them are being experimentally validated using a heterologous expression system. Taken together, the present work utilizes the power of Artificial Intelligence to deorphanize therapeutic-relevant GPCRs.