| アブストラクト | Gonadotropin-releasing hormone receptor (GnRHR) agonists are widely used therapeutically, yet their adverse event profile remains insufficiently characterized. We developed a machine-learning model using Tox21 GnRHR agonist activity data and molecular descriptors to predict GnRHR agonist activity among FDA Adverse-Event Reporting System (FAERS)-listed drugs. BalancedRandomForest achieved the highest ROC-AUC (0.810) and was applied to 5523 FAERS-listed drugs. Following applicability-domain assessment, 1191 drugs were retained, of which 367 were predicted to have GnRHR agonist activity. FAERS data from 2004 Q1 through 2024 Q3 were analyzed using reporting odds ratios (RORs) and Fisher's exact test for MedDRA Preferred Terms (PTs). Overall, 330 unique PTs corresponding to 531 PT-SOC assignments met prespecified criteria: at least 1000 reports, ln(ROR) > 1, and q < 0.05. The highest proportions of significant PTs were observed in respiratory, thoracic and mediastinal disorders; infections and infestations; and hepatobiliary disorders. These findings suggest that drugs predicted to have GnRHR agonist activity may show disproportionate reporting of respiratory, infectious, and hepatobiliary adverse events, although causality cannot be established. This framework may support hypothesis generation and safety signal prioritization during drug development and postmarketing surveillance. |
|---|
| ジャーナル名 | International journal of molecular sciences |
|---|
| Pubmed追加日 | 2026/8/13 |
|---|
| 投稿者 | Migura, Yui; Uesawa, Yoshihiro |
|---|
| 組織名 | Department of Medical Molecular Informatics, Meiji Pharmaceutical University,;Tokyo 204-8588, Japan.;Tokushima Research Center for Drug Discovery, Otsuka Pharmaceutical Co., Ltd.,;Tokushima 771-0192, Japan. |
|---|
| Pubmed リンク | https://www.ncbi.nlm.nih.gov/pubmed/42589332/ |
|---|
