非酒精性脂肪肝病（NAFLD）是当前未满足的医疗需求，FXR激动剂是防治NAFLD的潜在疗法。已知的FXR激动剂疗效欠佳或具有严重副作用，因而迫切需要研发新骨架FXR激动剂。基于结构的药物设计可加快新药发现，但蛋白质柔性问题制约了该技术的广泛应用。项目组前期发展了考虑蛋白质柔性的集成学习小分子药效构象识别方法，建立了FXR靶标特异性融合打分函数，并发现了口服有效的新骨架FXR激动剂XJ034，解析了复合物晶体结构。本项目拟进一步发展人工智能小分子药效构象识别方法，建立高精度的FXR靶标特异性融合打分函数，将其应用于基于结构的XJ034优化，通过化学合成、构效关系研究和体内外生物学活性评价，获得抗NAFLD药效和药代动力学性质优于XJ034的FXR激动剂。本项目将为精准药物设计提供一种新颖的基于人工智能小分子药效构象识别的打分函数，同时为防治NAFLD提供具有开发前景的药物先导物或候选物。

Non-alcoholic fatty liver disease (NAFLD) represents an unmet medical need at present. FXR agonists are potential therapeutics to prevent and treat NAFLD. The lack of efficacy or severe side effects of the current FXR agonists make it urgent to discover new chemotypes. Structure-based drug design can speed up new drug discovery, while its wide application is somewhat hindered by protein flexibility. Our research team previously developed an ensemble-learning-based approach to predict small-molecule bioactive conformations, which considered protein flexibility. Based on that approach, we built a FXR-specific fusion scoring function, which led to the discovery of XJ034 as an orally active, novel-scaffold FXR agonist. Furthermore, we determined the crystal structure of FXR in complex with XJ034. In this study, we propose to further develop Artificial Intelligence(AI)-based approaches to predict small-molecule bioactive conformations and FXR-specific fusion scoring functions of high accuracy. We will apply them to the structural optimization of XJ034 based on the crystal structure of FXR in complex with XJ034. By chemical synthesis, structure-activity relationship study as well as biological evaluation in vitro and in vivo, we aim to obtain new FXR agonists that show better anti-NAFLD efficacy and PK properties than XJ034. This study will provide a novel scoring function based on small-molecule bioactive conformations recognized by AI to facilitate precise drug design. It will also provide novel lead compounds or candiates for the prevention and treatment of NAFLD.