Harnessing an artificial intelligence-based approach to target vulnerabilities in uveal melanoma

Abstract Introduction Metastatic uveal melanoma (UM) is a deadly disease with a high propensity for liver metastasis. Despite recent advances, there remains an unmet need for effective therapies. Approximately 80-90% of UM cases are driven by activating mutations in GNAQ and GNA11, leading to aberrant signaling through Gαq subunit, a critical component of heterotrimeric G-protein. Computational drug screening is a powerful strategy enabling rapid identification of compounds modulating key signaling pathways. This study aims to identify novel drug candidates for UM targeting previously uncharacterized binding pockets in the oncogenic Gαq protein using iTripleD, an innovative in silico approach. Method The 3D structure of Gαq was retrieved from the Protein Data Bank to identify potential unique ligand-binding pockets. iTripleD was then used to screen Mcule and Molport compound libraries for drug candidates targeting Gαq in preclinical UM models. Result The Protein Data Bank revealed three binding pockets in Gαq: the guanine triphosphate (GTP) binding site, the Gα–Gβγ interface, and a unique effector binding site for which no compounds have previously been identified. iTripleD efficiently screened 37.5 million compounds at a rate of 2,000 molecules per second and identified 63 small-molecule inhibitors that met the standard pharmacological criteria and showed high predicted binding affinity to these three critical sites on the Gαq protein. Discussion In vitro studies are underway to characterize the impact of these inhibitors on downstream signaling pathways, which will be validated in in vivo xenograft mouse models of UM. These findings will help advance key compounds toward clinical translation in GNAQ/GNA11 mutant UM.