NDI-101150

点击蓝字 关注我们 引用本文：王超，丁晓，王亚洲，等．人工智能在抗肿瘤药物发现中的应用进展[J/CD]．肿瘤综合治疗电子杂志，2025，11（1）：36-43． 通信作者：任峰 E-mail ：feng.ren@insilico.ai 【摘要】 药物发现往往面临周期长、成本高、成功率低等问题，人工智能（artificial intelligence，AI）因算力、算法的不断更新迭代而飞速发展，为新药研发注入了强大动力。本文从靶点的发现与确证、蛋白质结构预测、先导化合物的发现与优化、分子性质预测以及加速临床试验等方面对AI在抗肿瘤药物发现中的应用进行简要介绍。随后，对AI参与的代表性抗肿瘤药物分子的临床进展进行了简要分析，以期为行业研究方向提供启发和参考。 【关键词】 人工智能；药物发现；抗肿瘤药物；临床进展 人工智能（artificial intelligence，AI）的概念始于1956年达特茅斯会议，计算机科学家们提出建立具有完全人类智能的新型计算机，从而诞生了使用计算技术来模拟、延伸人类思维过程和智能行为的新技术科学[1]。至2010年后，深度神经网络在图像和语音识别等领域表现出色，AI技术已经在医疗影像、自然语言处理、医用材料等各个领域发挥着越来越重要的作用，已从理论转变为具有解决复杂问题能力的实际应用工具[2-6]。 2024年，瑞典皇家科学院将诺贝尔物理学奖授予科学家John J. Hopfield和Geoffrey E. Hinton，以表彰他们为推动利用人工神经网络进行机器学习作出的基础性发现和发明[7-8]。同时，诺贝尔化学奖颁给了科学家Demis Hassabis和John M. Jumper，表彰他们在计算蛋白质设计和蛋白质结构预测方面的贡献[9-11]。这2项科学研究的最高奖项更是将AI应用于生物医药的关注推向了新的高潮。 在新药研发领域，众多跨国药企拥抱AI技术。全球疫苗市场中的四大巨头之一的赛诺菲提出“All in AI”战略，与Exscientia、Formation Bio、OpenAI、Insilico Medicine等AI公司建立合作，将数据、软件和经过调整的模型汇于一体，共同优化全周期的方案。阿斯利康利用Owkin庞大的数据网络，从数字化病理切片中优化胚系乳腺癌易感基因（germline breast cancer susceptibility gene，gBRCA）突变的预筛查，同时扩大与Immunai公司的合作，在癌症免疫疗法临床试验设计中使用AI驱动的免疫细胞图谱。诺华与药物研发平台Generate Biomedicine达成多靶点合作，以生成式AI发现和开发蛋白质疗法；最近又与分子胶AI公司Monte Rosa Therapeutics达成协议，推进靶向鸟嘌呤核苷酸交换因子1（vav guanine nucleotide exchange factor 1，VAV1）的分子胶蛋白降解剂的开发。礼来公司则与生物技术公司Genetic Leap达成合作协议，利用其AI平台寻找RNA靶向小分子药物。AI制药赛道的其他更多投融资和合作，表明AI对复杂制药领域的实际价值。本文旨在对AI在药物研发中的着力点进行简要介绍，同时通过简要分析AI参与的代表性抗肿瘤药物临床进展，以期为行业研究方向提供启发和参考。 1 人工智能在药物研发中的应用概述 基于近年来算力、算法的不断发展，AI在制药领域发挥着越来越重要的作用[12-15]。AI应用主要包括了药物靶点发现与确证、蛋白结构预测及生成、先导化合物的发现与优化、分子性质预测以及加速临床试验等（图1）。 1.1 靶点发现与确证 基于靶点的药物发现是药物研发的主流手段，药物靶点对整个新药研发项目起到决定性的作用[16]。系统生物学通过研究各个生物系统内部各成分间相互关系，建立整个系统的可理解模型，为有机体绘制完整图谱[17-19]。AI生物技术公司BERG开发的Interrogative Biology是一个基于AI的系统生物学平台，可生成数据驱动的无偏网络，用于识别靶点和疾病的生物标志物[20]。BenevolentAI推出的判断加强认知系统通过发现疾病、药物、试验数据等大量非结构化数据间的新联系，实现药物重定向（老药新用）[21]。Insilico Medicine的科学家基于信号通路激活分析提出了通路激活网络分解分析的方法，可以从大规模转录组学和蛋白质组学数据中提取生物学相关特征[22]。 1.2 蛋白结构预测及生成 截至目前，针对蛋白结构预测的算法大致可分为同源建模、从头建模和基于机器学习建模3类[23]。基于深度学习的建模方法是一种数据驱动方法，其中AlphaFold[10]、RoseTTAFold[11]、ESMFold[24]和AminoBERT[25]等语言模型最具代表性。作为最先进的蛋白结构预测方法，AlphaFold2能以原子级精度预测蛋白质结构，并能大规模运行，这一重大突破展示了前所未有的应用场景[26]。进一步升级后的AlphaFold3则可更快、更高精度地预测核酸（DNA/RNA）和更小分子的3D结构，并揭示它们之间的相互作用以及预测结合模式，为活性配体的发现提供了更加广阔的前景[27]。除了结构预测之外，利用扩散模型对蛋白质进行生成也取得了重要进展。蛋白质设计先驱David Baker教授利用RFdiffusion能生成各种功能性蛋白质，包括在天然蛋白质中从未见过的拓扑结构[28]。Generate Biomedicines可利用扩散模型工具Chroma对蛋白质进行从头生成的工作，进一步拓展了蛋白质结构[29]。 1.3 先导化合物发现及优化 基于深度学习的生成模型可从大量药物数据中学习化学分子结构与其生物和药理性质之间的非线性概率分布，然后根据所需性质，重新生成分子[30-33]。对于生成式建模，目前常用的是基于简化分子线性输入规范的字符串编码算法，代表性的研究包括基于分子图的生成对抗网络[34]、目标增强生成对抗网络[35]、对抗式自编码器[36]、高斯混合模型变分自编码器[37]及生成张量强化学习模型[38]等优化方法。 1.4 分子性质预测 以机器学习、深度学习为代表的药物吸收、分配、代谢、排泄和毒性（absorption，distribution，metabolism，excretion，toxicity，ADMET）预测模型可从数据中学习化学结构与药代动力学性质的关联，避免不良未知空间的探索[39-40]。目前已有不少AI软件可用于ADMET性质的预测，例如，Simulation Plus公司的GastroPlus已经被多国药监部门广泛使用，可预测多个药代动力学参数以及药物经眼部、肺部等多个途径给药后的体内过程等。拜耳医药研究人员总结其ADMET计算工具时表示这些工具的成功主要取决于模型质量、与研究过程的模型相关性以及结果的可获取和解释性[41]。此外，药物晶型会对溶解度、溶出度、稳定性、疗效等产生影响，通过计算模拟进行晶型预测对保障药物研发的速度和质量具有重要意义[42]。晶态科技在与AbbVie公司合作的项目中将晶体堆积作用考虑进算法模型，可对具有相似骨架的同系列分子进行溶解度区分[43]。 1.5 加速临床试验 众所周知，临床试验的高失败率导致药物研发的高成本和低效率，在药物研发后期于临床试验设计阶段的关键步骤中使用AI，有助于实现更精准的患者分层，提高招募效率，从而提高临床试验的成功率[44-45]。Insilico Medicine于2022年推出可预测临床试验的inClinico工具，其是一个基于Transformer模型的AI系统，预测临床试验Ⅱ～Ⅲ期的转化成功率[46]。inClinico整合了多种评分方法、多模态数据源（包括文本、组学、临床试验设计和小分子特性）以及生物学背景，并与最先进的深度学习模型相结合。在前瞻性验证中，inClinico在预测临床试验Ⅱ～Ⅲ期的转化成功率的准确率达到了79%。 2 人工智能参与的临床药物进展 AI在药物研发的整个流程中均已有不同程度的应用，目前已有不少AI参与研发的药物分子进入临床阶段（表1）[47-48]。接下来将对代表性药物分子的临床进展情况进行简要分析。 早期AI参与的部分临床药物来自对传统药物分子的适应证拓展。例如，REC-2282是最早由俄亥俄州立大学研究开发的口服小分子HDAC抑制剂AR-42，早期临床试验显示，REC-2282在接受肿瘤切除术的成年患者前庭神经鞘瘤中抑制了异常的AKT、S6通路激活（NCT02282917）[49-50]。Recursion公司利用Recursion OS系统发现了REC-2282用于2型神经纤维瘤（neurofibromatosis type 2，NF2）的治疗潜力，并借助细胞和动物模型证明了这一发现。目前该分子正在开展针对进展性NF2脑膜瘤患者的Ⅱ/Ⅲ期临床试验（NCT05130866）。REC-4881是Recursion公司从武田制药公司引进的口服小分子MEK1/2变构抑制剂，Recursion公司利用Recursion OS平台发现特异性MEK1/2抑制剂在APC突变的多发性肠肿瘤小鼠模型中显示息肉和异常增生腺瘤的减少，目前REC-4881正在开展治疗家族性腺瘤息肉病（familial adenomatous polyposis，FAP）的临床Ⅰ/Ⅱ期研究（NCT05552755），于2022年7月获得欧盟委员会授予的用于潜在治疗FAP的孤儿药资格认证[51]。 AI参与设计的全新结构的药物分子逐渐进入临床研究。RLY-4008是Relay公司基于蛋白动力学设计的FGFR-2选择性抑制剂，在FGFR-2改变的胆管癌患者的临床试验中（NCT04526106），全部剂量统计下（n＝38）的客观缓解率（objective response rate，ORR）为57.9%，在推荐的Ⅱ期剂量70 mg/d的患者中（n＝17）ORR可达82.4%，明显优于获批的泛FGFR抑制剂。RLY-4008具有可控制的毒性特征，常见的3级治疗相关不良反应（n＝124）包括口腔炎（11%）、手足综合征（7%）、指甲毒性（3%）、疲劳（1%）以及视网膜色素上皮脱离（1%）[52-54]。NDI-101150为Nimbus公司通过基于结构的药物发现引擎研发的HPK1抑制剂，在临床前研究中显示了对同家族激酶＞ 300倍的高选择性[55]。在开展的临床Ⅰ/Ⅱ期试验中评估其作为单一疗法（50～200 mg剂量）或与帕博利珠单抗（200 mg）联合治疗晚期实体瘤（NCT05128487）。在已公布的临床数据中，16.7%（4/24）的患者临床获益，其中1例有肾细胞癌的患者获得了完全应答，3例患者维持了持续稳定的疾病超过6个月[56]。MRT-2359是Monte Rosa公司借助QuEEN平台设计研发的口服GSPT1分子胶降解剂，在体外试验和体内模型中均显示了良好的GSPT1降解以及抗肿瘤活性[57-58]。研究发现，MRT-2359在多种癌症细胞系中表现出显著的抗增殖活性，尤其在Myc驱动的细胞系中。在2023年公布的临床Ⅰ/Ⅱ期数据（NCT05546268） 中显示，对21例给药患者中的15例进行疗效评估中，0.5～2.0 mg的给药区间均观察到外周血单核细胞和肿瘤组织活检GSPT1蛋白表达减少约60%。在6例生物标志物呈阳性的患者中，2例实现了部分缓解，1例达到了病情稳定。0.5 mg和1.0 mg剂量下仅观察到1级或2级治疗相关不良事件[59]。研究人员认为0.5 mg剂量下，给药21 d停药7 d的方案是潜在的推荐Ⅱ期剂量，目前正在进行0.75 mg剂量评估。 AI不仅在口服小分子药物的生成和优化中发挥了重要作用，也可应用于生物大分子的设计优化[60]。Aulos Bioscience公司通过机器学习发现特定突变位点，设计研发了特异性阻断IL-2与CD25结合的高度亲和性抗体AU-007，这是首款进入临床试验的AI大分子药物。AU-007于2022年进入临床研究，在2024年美国临床肿瘤学会年会上公布了积极的临床Ⅰ/Ⅱ期数据（NCT05267626）[61-62]。研究表明，在临床早期阶段即发现了AU-007具有显著的抗肿瘤活性，同时在部分患者中观察到肿瘤减少的持久性，表明形成了癌细胞的免疫记忆。与此同时，该药物显示了良好的耐受性（n＝53），除4例患者出现短暂性淋巴细胞减少（3级和4级）和1例患者出现3级贫血外，所有药物相关不良反应均为1级或2级，最常见的包括发热（18%）、疲劳（16%）、恶心（14%）、淋巴细胞减少（8%）和寒战（6%）。该治疗方法克服了由其他白细胞介素-2治疗引起的负反馈回路，并使免疫系统偏向激活，可能是潜在的最佳治疗方案，为AI辅助大分子药物设计打开了全新的窗口。 临床阶段除了上述正在进行的分子，也有停止进展的案例。例如，EXS-21546是由Exscientia利用AI驱动平台和Evotec公司共同研发的腺苷A2A受体拮抗剂，在临床Ⅰ期的健康受试者的药代动力学（pharmacokinetics，PK）/药效学（pharmacodynamics，PD）及安全性研究中，EXS-21546在CD8阳性细胞中对环磷酸腺苷效应元件结合蛋白磷酸化表现出剂量依赖性抑制，PD特性可反映血浆暴露量，A2A受体信号转导的抑制作用在2次/d给药期间持续存在。同时，在单次给药剂量递增全部剂量以及多次给药剂量递增的150 mg 2次/d剂量下，未报告中枢神经系统不良反应（NCT04727138）。随后，在复发或难治性肾细胞癌和非小细胞肺癌患者中开展了与程序性死亡受体1抑制剂联用的Ⅰ/Ⅱ期临床研究。遗憾的是，2023年10月因疗效未达预期而终止了该临床研究（NCT05920408）[63-64]。目前，Exscientia管线中进展最快的药物是与GT Apeiron合作设计的CDK7抑制剂GTAEXS617，该药物正在进行临床Ⅰ/Ⅱ期测试，用于治疗晚期实体瘤，有望成为耐药性周期蛋白依赖性激酶（cyclin dependent kinases，CDK）4/6抑制剂有潜力的替代选择（NCT05985655）[65]。 其他一些处于临床Ⅰ期阶段的药物分子，在临床前研究中显示了较好的抗肿瘤活性，目前尚无临床数据报道。XL309（ISM3091）是一款靶向USP1的潜在同类最佳小分子抑制剂，由Insilico Medicine自主研发的AI平台赋能开发，在BRCA突变的肿瘤模型中单药治疗或与靶向聚ADP核糖聚合酶抑制剂奥拉帕尼联合使用，均展示了强效的抗肿瘤活性[66]。Insilico Medicine授予Exelixis开发和商业化ISM3091及其他靶向USP1化合物的全球独家许可，于2023年第3季度获得8000万美元预付款，正在进行临床Ⅰ期研究（NCT05932862）。Schrödinger公司开发的MALT1蛋白酶抑制剂SGR-1505在体外酶活与细胞测试以及与B淋巴细胞瘤-2抑制剂维奈克拉的联用试验中均表现出优于临床分子JNJ-6633的活性，同时，对核因子-κB及相关信号通路的基因表达有更大的影响，目前正在针对B细胞恶性肿瘤患者进行临床Ⅰ期试验（NCT05544019）[67]。SGR-3515为Schrödinger公司研发的MYT1/WEE1（WEE家族蛋白激酶）共抑制剂，在临床前研究中对WEE1表现出强结合慢解离特性，同时特异性肿瘤组织分布，在A427和OVCAR3肿瘤模型中支持给药3 d停药11 d的间歇给药方案，减少潜在的骨髓抑制和消化道毒性。由于WEE1与MYT1的协同效应，SGR-3515显示出优于单独抑制WEE1或MYT1的抗肿瘤活性。目前，SGR-3515正在进行晚期实体瘤的Ⅰ期剂量递增研究（NCT06463340）[68]。PHI-101是Pharos iBio公司利用其基于大数据的AI平台Chemi-nverse研发的新一代FLT3小分子抑制剂。临床前研究表明，PHI-101有望克服多种耐药突变产生的耐药性，目前，在开展针对急性髓性白血病和高级别浆液性卵巢癌患者的临床Ⅰ期研究（NCT04842370/NCT04678102）[69-70]。ISN018_055是Insilico Medicine利用“端到端”药物开发平台Pharma.AI发现的Traf2和Nck相互作用激酶小分子抑制剂，AI深度赋能靶点发现、分子生成和临床方案设计全过程。2024年9月，Insilico Medicine公布了ISN018_055临床Ⅱa期积极结果，显示该药物具有良好的安全性和剂量依赖性的药效，是AI制药领域的里程碑[71-72]。 3 总结与展望 新药的研发成本和周期不断提高，而成功率却不断下降，很多疾病仍得不到有效治疗，AI技术的发展无疑为这些困境带来突破口和强劲动力[73]。在传统药物分子设计存在挑战的领域，AI的参与将带来新的契机。例如，复杂靶点如G蛋白偶联受体和RNA等，其与配体的结合模式往往与常规蛋白有所不同；分子胶降解剂涉及蛋白-蛋白界面作用和探索全新口袋，三元复合物存在稳定性问题，给结合模式预测及分子生成带来较大难度[74-78]。传统中药作为我国千百年来实践经验的结晶，在治疗棘手疾病方面展现了巨大潜力，AI技术也已有所涉猎，例如，新的组合处方、药理机制、网络药理学、患者数据的相关性分析等[79-80]。大语言模型、扩散模型、多任务学习等新兴技术在这些具有挑战性的药物发现领域也具有广阔的前景[81-84]。 高质量的数据是AI模型的基础，但制药领域的数据往往很有限，非常分散且格式多样，需要大量的清洗和整理工作[85]。另外面临负样本缺失的问题，缺少失败实验的相关数据，会导致机器学习模型中的协变量偏移，从而影响分类精度。随着AI技术的不断进步，需要保障数据隐私、算法透明度以及公平性等问题。通过实验室信息管理系统和电子实验记录本等方式，有望实现实验数据的有效管理和共享[86]。虽然AI技术可在不同阶段助力药物的发现与开发，但新药发现是个复杂且庞大的工程，整体而言仍需持谨慎乐观态度。一方面AI助力药物研发大多停留在早期发现阶段；另一方面进入临床的AI药物大部分仍在临床Ⅰ/Ⅱ期阶段，整体评价下药物安全性和PK/PD性质较良好，而在关键的Ⅱ期和Ⅲ期有效性验证中则表现不一，尚缺乏足够样本的有效性验证。除上述分析的抗肿瘤药物外，针对非肿瘤靶点的部分AI药物分子进展不如预期，例如，由Exscientia与住友制药共同开发的用于治疗强迫症的长效5-羟色胺1A受体激动剂DSP-1181，由于临床Ⅰ期研究未达到预期标准已停止开发。住友制药和大冢制药合作的精神分裂症药物ulotaront，以及BenevolentAI公司的特应性皮炎药物BEN-2293均由于药效未达预期终点而在临床中后期停止研究。 近期2家诞生于AI制药初期的公司Recursion和Exscientia宣布合并，也启发了AI制药的商业模式[87]。随着高质量数据的不断累积、AI技术平台的不断优化以及AI制药公司商业模式的持续探索，未来会有更多临床真实数据帮助检验AI药物的安全性和有效性。相信未来AI药物将实现概念验证并获批上市，帮助解决未满足的临床需求，全面加速新药研发。 参考文献 向上滑动查看 [1] MCCARTHY J. Artificial intelligence, logic and formalizing common sense[M]. Dordrecht: Springer, 1989:161-190. [2] YAN L, GREIFF S, TEUBER Z, et al. Promises and challenges of generative artificial intelligence for human learning[J]. Nat Hum Behav, 2024, 8(10):1839-1850. [3] TRAN H, GURNANI R, KIM C, et al. Design of functional and sustainable polymers assisted by artificial intelligence[J]. Nat Rev Mater, 2024, 9:866-886. [4] RAJPURKAR P, CHEN E, BANERJEE O, et al. AI in health and medicine[J]. Nat Med, 2022, 28(1):31-38. [5] PEREZ-LOPEZ R, LALEH N G, MAHMOOD F, et al. A guide to artificial intelligence for cancer researchers[J]. Nat Rev Cancer, 2024, 24(6):427-441. [6] WANG H, FU T, DU Y, et al. Scientific discovery in the age of artificial intelligence[J]. Nature, 2023, 620(7972):47-60. [7] GIBNEY E, CASTELVECCHI D. Physics Nobel scooped by machine-learning pioneers[J]. Nature, 2024, 634(8034):523-524. [8] LECUN Y, BENGIO Y, HINTON G. Deep learning[J]. Nature, 2015, 521(7553):436-444. [9] CALLAWAY E. Chemistry nobel goes to developers of AlphaFold AI that predicts protein structures[J]. Nature, 2024, 634(8034):525-526. [10] JUMPER J, EVANS R, PRITZEL A, et al. Highly accurate protein structure prediction with AlphaFold[J]. Nature, 2021, 596(7873):583-589. [11] BAEK M, DIMAIO F, ANISHCHENKO, et al. Accurate prediction of protein structures and interactions using a three-track neural network[J]. Science, 2021, 373(6557):871-876. [12] YANG X, WANG Y, BYRNE R, et al. Concepts of artificial intelligence for computer-assisted drug discovery[J]. Chem Rev, 2019, 119(18):10520-10594. [13] JAYATUNGA M K P, XIE W, RUDER L, et al. AI in small-molecule drug discovery: a coming wave?[J]. Nat Rev Drug Discov, 2022, 21(3):175-176. [14] XU W. Current status of computational approaches for small molecule drug discovery[J]. J Med Chen, 2024, 67(21):18633- 18636. [15] CATACUTAN D B, ALEXANDER J, ARNOLD A, et al. Machine learning in preclinical drug discovery[J]. Nat Chem Biol, 2024, 20(8):960-973. [16] SANTOS R, URSU O, GAULTON A, et al. A comprehensive map of molecular drug targets[J]. Nar Rev Drug Discov, 2017, 16(1):19-34. [17] ZUPANIC A, BERNSTEIN H C, HEILAND I. Systems biology: current status and challenges[J]. Cell Mol Life Sci, 2020, 77(3):379-380. [18] ALQURAISHI M, SORGER P K. Differentiable biology: using deep learning for biophysics-based and data-driven modeling of molecular mechanisms[J]. Nat Methods, 2021, 18(10):1169- 1180. [19] ZHOU R, TANG X, WANG Y. Emerging strategies to investigate the biology of early cancer[J]. Nat Rev Cancer, 2024, 24(12):850-866. [20] NARAIN N, KIEBISH M, VISHNUDAS V, et al. CSAO-1. Interrogative biology: unraveling insights into causal disease drivers by use of a dynamic systems biology and Bayesian AI to identify the intersect of disease and healthy signatures[J]. Neuro Oncol Adv, 2021, 3(Suppl 2):ii1. [21] RICHARDSON P, GRIFFIN I, TUCKER C, et al. Baricitinib as potential treatment for 2019-nCoV acute respiratory disease[J]. Lancet, 2020, 395(10223):e30-e31. [22] OZEROV I V, LEZHNINA K V, IZUMCHENKO E, et al. In silico pathway activation network decomposition analysis (iPANDA) as a method for biomarker development[J]. Nat Commun, 2016, 7(1):1-11. [23] ALQURAISHI M. Machine learning in protein structure prediction[J]. Curr Opin Chem Biol, 2021, 65:1-8. [24] LIN Z M, AKIN H, RAO R, et al. Evolutionary-scale prediction of atomic-level protein structure with a language model[J]. Science, 2023, 379(6637):1123-1130. [25] CHOWDHURY R, BOUATTA N, BISWAS S, et al. Single-sequence protein structure prediction using a language model and deep learning[J]. Nat Biotechnol, 2022, 40(11):1617-1623. [26] YANG Z Y, ZENG X X, ZHAO Y, et al. AlphaFold2 and its applications in the fields of biology and medicine[J]. Signal Transduct Target Ther, 2023, 8(1):115. [27] ABRAMSON J, ADLER J, DUNGER J, et al. Accurate structure prediction of biomolecular interactions with AlphaFold3[J]. Nature, 2024, 630(8016):493-500. [28] WATSON J L, JUERGENS D, BENNETT N R, et al. De novo design of protein structure and function with RFdiffusion[J]. Nature, 2023, 620(7976):1089-1100. [29] INGRAHAM J B, BARANOV M, COSTELLO Z, et al. Illuminating protein space with a programmable generative model[J]. Nature, 2023, 623(7989):1070-1078. [30] VANHAELEN Q, LIN Y C, ZHAVORONKOV A. The advent of generative chemistry[J]. ACS Med Chem Lett, 2020, 11(8):1496-1505. [31] IVANENKOV Y, ZAGRIBELNYY B, MALYSHEV A, et al. The hitchhiker's guide to deep learning driven generative chemistry[J]. ACS Med Chem Lett, 2023, 14(7):901-915. [32] DU Y, JAMASB A R, GUO J, et al. Machine learning-aided generative molecular design[J]. Nat Mach Intell, 2024, 6(6):589-604. [33] KENAKIN T. Know your molecule: pharmacological characteriz-ation of drug candidates to enhance efficacy and reduce late-stage attrition[J]. Nat Rev Drug Discov, 2024, 23(8):626-644. [34] GOODFELLOW I J, POUGET-ABADIE, MIRZA M, et al. Generative adversarial nets[J]. Adv Neural Inf Process Syst, 2014:2672-2680. [35] GUIMARAES G, SANCHEZ-LENGELING B, OUTEIRAL C, et al. Objective-reinforced generative adversarial networks (ORGAN) for sequence generation models[J]. arXiv, 2017: 1705.10843. [36] KADURIN A, ALIPER A, KAZENNOV A, et al. The cornu copia of meaningful leads: applying deep adversarial autoen coders for new molecule development in oncology[J]. Oncotarget, 2017, 8(7):10883-10890. [37] YU Y, XU T Y, LI J W, et al. A novel scalarized scaffold hopping algorithm with graph-based variational autoencoder for discovery of JAK1 inhibitors[J]. ACS Omega, 2021, 6(35): 22945-22954. [38] ZHAVORONKOV A, IVANENKOV Y A, ALIPER A, et al. Deep learning enables rapid identification of potent DDR1 kinase inhibitors[J]. Nat Biotechnol, 2019, 37(9):1038-1040. [39] BHATIA A S, SAGGI M K, KAIS S. Quantum machine learning predicting ADME-Tox properties in drug discovery[J]. J Chem Inf Model, 2023, 63(21):6467-6486. [40] BHHATARAI B, WALTERS W P, HOP C E, et al. Opportu-nities and challenges using artificial intelligence in ADME/ Tox[J]. Nat Mater, 2019, 18(5):418-422. [41] GOLLER A H, KUHNKE L, MONTANARI F, et al. Bayer's in silico ADMET platform: a journey of machine learning over the past two decades[J]. Drug Discov Today, 2020, 25(9):1702-1709. [42] XIOURAS C, CAMELI F, QUILLO G L, et al. Applications of artificial intelligence and machine learning algorithms to crystallization[J]. Chem Rev, 2022, 122(15):13006-13402. [43] HONG R S, MATTEI A, SHEIKH A Y, et al. Novel physics-based ensemble modeling approach that utilizes 3D molecular conformation and packing to access aqueous thermodynamic solubility: a case study of orally available bromodomain and extraterminal domain inhibitor lead optimization series[J]. J Chem Inf Model, 2021, 61(3):1412-1426. [44] JAYATUNGA M K, AYERS M, BRUENS L, et al. How successful are AI-discovered drugs in clinical trials? A first analysis and emerging lessons[J]. Drug Discov Today, 2024, 29(6):104009. [45] SONG A H, JAUME G, WILLIAMSON D F K, et al. Artificial intelligence for digital and computational pathology[J]. Nat Rev Bioeng, 2023, 1(12):930-949. [46] ALIPER A, KUDRIN R, POLYKOVSKIY D, et al. Prediction of clinical trials outcomes based on target choice and clinical trial design with multi-modal artificial intelligence[J]. Clin Pharmacol Ther, 2023, 114(5):972-980. [47] ZHANG Y, MASTOURI M, ZHANG Y. Accelerating drug discovery, development, and clinical trials by artificial intelli-gence[J]. Med, 2024, 5(9):1050-1070. [48] ARNOLD C. Inside the nascent industry of AI-designed drugs[J]. Nat Med, 2023, 29(6):1292-1295. [49] WELLING D B, COLLIER K A, BURNS S S, et al. Early phase clinical studies of AR-42, a histone deacetylase inhibitor, for neurofibromatosis type 2-associated vestibular schwann-omas and meningiomas[J]. Laryngoscope Investig Otolaryngol, 2021, 6(5):1008-1019. [50] COLLIER K A, VALENCIA H, NEWTON H, et al. A phase 1 trial of the histone deacetylase inhibitor AR-42 in patients with neurofibromatosis type 2-associated tumors and advanced solid malignancies[J]. Cancer Chemother Pharmacol, 2021, 87(5):599-611. [51] ADJEI A A, LORUSSO P, RIBAS A, et al. A phase Ⅰ dose-escalation study of TAK-733, an investigational oral MEK inhibitor, in patients with advanced solid tumors[J]. Invest New Drugs, 2017, 35(1):47-58. [52] SCHONHERR H, AYZA P, TAYLOR A M, et al. Discovery of lirafugratinib (RLY-4008), a highly selective irreversible small-molecule inhibitor of FGFR2[J]. Proc Natl Acad Sci U S A, 2024, 121(6):e2317756121. [53] SUBBIAH V, SAHAI V, MAGLIC D, et al. RLY-4008, the first highly selective FGFR2 inhibitor with activity across FGFR2 alterations and resistance mutations[J]. Cancer Discov, 2023, 13(9):2012-2031. [54] SCHRAM A M, SUBBIAH V, LIAN C Y, et al. Clinical activity of lirafugratinib (RLY-4008), a highly selective FGFR2 inhibitor, in patients with advanced FGFR2-altered solid tumors: the ReFocus study[J]. Mol Cancer Ther, 2023, 22 (Suppl 12):IA006. [55] CICCONE D, KUO F S, BOIKO S, et al. NDI-101150 is a potent and highly selective hematopoietic progenitor kinase 1 (HPK1) inhibitor that promotes a robust and broad anti-tumor immune response[J]. J Immunother Cancer, 2023, 11(Suppl 1): A1494. [56] NOEL M S, DEMEL K C, SRIVASTAVA B, et al. Phase 1/2 trial of the HPK1 inhibitor NDI-101150 as monotherapy and in combination with pembrolizumab: clinical update[J]. J Clin Oncol, 2024, 42(Suppl 16):3083. [57] WALLACE O B. Discovery of MRT-2359, an orally bioav-ai lable GSPT1 molecular glue degrader, for MYC-driven cancers[J]. Cancer Res, 2023, 83(Suppl 7):3449. [58] TIEDT R, SCHILLO M, OSMONT A, et al. The GSPT1 molecular glue degrader MRT-2359 is active against prostate cancer[J]. Cancer Res, 2024, 84 (Suppl 6):3294. [59] Biospace. Monte rosa therapeutics announces interim PK/PD and clinical data for MRT-2359 in phase 1/2 trial for MYC-Driven solid tumors[EB/OL]. (2023-10-17)[2024-11-26]. https://www.biospace.com/monte-rosa-therapeutics-announces-interim-pk-pd-and-clinical-data-for-mrt-2359-in-phase-1-2- trial-for-myc-driven-solid-tumors. [60] NAGRA N S, VEKEN L, STANZL E, et al. The company landscape for artificial intelligence in large-molecule drug discovery[J]. Nat Rev Drug Discov, 2023, 22(12):949-950. [61] DOLGIN E. IL-2 upgrades show promise at ASCO[J]. Nat Biotechnol, 2022, 40(7):986-988. [62] RASCO D W, MCKEAN M, HAYDON A M, et al. Updated results of phase 1/2 study of AU-007, a monoclonal antibody (mAb) that binds to IL-2 and inhibits CD25 binding, in patients with advanced solid tumors[J]. J Clin Oncol, 2024, 42(Suppl 16):2527. [63] PAYNE A, FONS P, ALT I, et al. Abstract 1731: EXS21546, a non-CNS penetrant A2AR-selective antagonist for anti-cancer immunotherapy[J]. Cancer Res, 2021, 81(Suppl 13):1731. [64] DIAZ N, GARRATT H, ZEMLA-BROWN A, et al. Abstract CT114: Data from first-in-human study of EXS21546, an A2A receptor antagonist, now progressing into phase 1 in RCC/ NSCLC[J]. Cancer Res, 2023, 83(Suppl 8):CT114. [65] BESNARD J, JOSEPH J, BAI H, et al. Abstract 3930: AI-driven discovery and profiling of GTAEXS-617, a selective and highly potent inhibitor of CDK7[J]. Cancer Res, 2022, 82(Suppl 12):3930. [66] LI Y, WU J, LIU J, et al. Abstract 502: ISM3091, a novel selective USP1 inhibitor as a targeted anticancer therapy[J]. Cancer Res, 2023, 83(Suppl 7):502. [67] YIN W, YE M, NARSHALL N, et al. SGR-1505 is a potent MALT1 protease inhibitor with a potential best-in-class profile[J]. Blood, 2023, 142(Suppl 1):2997. [68] SUN S, SILVERGLEID S, GRAY F, et al. 147 (PB135): Discovery of SGR-3515, a first-in-class Wee1/Myt1 inhibitor with differentiated pharmacological properties in xenograft tumor models[J]. Eur J Cancer, 2024, 211(Suppl 1):114669. [69] NAM K Y, IM J, HAN J, et al. PHI-101 is a potent third-generation FLT3 inhibitor developed to overcome resistance in acute myeloid leukemia[J]. Blood, 2020, 136(Suppl 1):28. [70] YOON S S, SHIN D Y, LEE J H, et al. A potent small molecule inhibitor of FLT3, PHI-101 overcomes resistance in acute myeloid leukemia: efficacy and PK/PD profile in phase 1 first in human study[J]. Blood, 2022, 140(Suppl 1):3365-3366. [71] REN F, ALIPER A, CHEN J, et al. A small-molecule TNIK inhibitor targets fibrosis in preclinical and clinical models[J]. Nat Biotechnol, 2025, 43(1):63-75. [72] EWALD C Y, PULOUS F E, LOK S W, et al. TNIK's emerging role in cancer, metabolism, and age-related diseases[J]. Trends Pharmacol Sci, 2024, 45(6):478-489. [73] DYBEKOV A V, KATRITCH V. Computational approached stream lining drug discovery[J]. Nature, 2023, 616(7958):673-685. [74] LI H, SUN X, CUI W, et al. Computational drug development for membrane protein targets[J]. Nat Biotechnol, 2024, 42(2):229-242. [75] CHEN W, SONG C, LENG L, et al. The application of artificial intelligence accelerates G protein-coupled receptor ligand discovery[J]. Engineering, 2024, 32:18-28. [76] KOVACHKA S, PANOSETTI M, GRIMALDI B, et al. Small molecule approached to targeting RNA[J]. Nat Rev Chem, 2024, 8(2):120-135. [77] ZHANG H, ZHANG L, LIN A, et al. Algorithm for optimized mRNA design improves stability and immunogenicity[J]. Nature, 2023, 621(7978):396-403. [78] GUNASINGHE K K, GINJOM I R, SAN H S, et al. Can current molecular accurately predict RNA inhibitors?[J]. J Chem Inf Model, 2024, 64(15):5954-5963. [79] PAN D, GUO Y, FAN Y, et al. Development and application of traditional Chinese medicine using AI machine learning and deep learning strategies[J]. Am J Chin Med, 2024, 52(3):605-623. [80] SONG Z, CHEN G, CHEN C Y C. AI empowering traditional Chinese medicine?[J]. Chem Sci, 2024, 15(41):16844. [81] ANAHAN M, MCDONELL K, REYNOLDS L. Role play with large language models[J]. Nature, 2023, 623(7987):493-498. [82] LTAUFDERHEIDE J, RANISCH R. The ethics of ChatGPT in medicine and healthcare: a systematic review on Large Language Models (LLMs)[J]. NPJ Digit Med, 2024, 7(1):183. [83] YANG L, ZHANG Z, SONG Y, et al. Diffusion models: a comprehensive survey of methods and applications[J]. arXiv, 2024:2209.00796. [84] ALLENSPACH S, HISS J A, SCHNEIDER G. Neural multi-task learning in drug design[J]. Nat Mach Intell, 2024, 6(2): 124-137. [85] RANT G, BOYLES F, BIRCHALL K, et al. The future of machine learning for small-molecule drug discovery will be driven by data[J]. Nat Comput Sci, 2024, 4(10):735-743. [86] EDFELDT K, EDWARDS A M, ENGKVIST O, et al. A data science roadmap for open science organizations engaged in early-stage drug discovery[J]. Nat Commun, 2024, 15(1):5640. [87] MULLARD A. Creating an AI-first drug discovery engine[J]. Nat Rev Drug Discov, 2024, 23(10):734-735. 声明：本文由“肿瘤界”整理与汇编，欢迎分享转载，如需使用本文内容，请务必注明出处。 编辑：lagertha 审核：松月

BOSTON--( BUSINESS WIRE )-- Nimbus Therapeutics , LLC ("Nimbus Therapeutics" or "Nimbus"), a biotechnology company that designs and develops breakthrough medicines for patients through its powerful computational drug discovery engine, today announced the advancement of its clinical oncology therapeutic pipeline, and the appointment of Peter J. Tummino, Ph.D., to President of Research and Development. Dr. Tummino will be responsible for advancing the company’s product portfolio through all phases of discovery and clinical development. The company announced the completion of an investigational new drug (IND) application submission for NDI-219216, a novel Werner syndrome helicase (WRN) inhibitor in development for the treatment of microsatellite instability high (MSI-H) tumors. Nimbus plans to initiate the first clinical trial of NDI-219216 in the first half of 2025 under the leadership of Anita Scheuber, M.D., Ph.D., Senior Vice President, Therapeutic Area Head, Oncology. "We are encouraged by the building momentum in our oncology portfolio including important advances in our non-covalent WRN inhibitor program, with data indicating NDI-219216 has the potential to be a best-in-class agent," said Dr. Tummino. "NDI-219216 has demonstrated compelling preclinical data, including significant tumor regression and complete responses at low oral doses across tumor types. The preclinical safety studies suggest a promising benefit-risk profile as we prepare to evaluate this compound in patients. We are excited to advance NDI-219216 into a first-in-human clinical trial later this year," said Dr. Scheuber. The company recently completed its Phase 1/2 clinical study of NDI-101150, a highly selective and potent hematopoietic progenitor kinase 1 (HPK1) inhibitor, for the treatment of advanced solid tumors. Clinical data presented in November 2024 at the Society for Immunotherapy of Cancer (SITC) 39 th Annual Meeting demonstrated monotherapy efficacy with a favorable safety profile that supports potential combination therapy approaches. The company is actively evaluating opportunities to maximize the therapeutic potential of this program. Nimbus continues to progress drug discovery efforts across metabolic and autoimmune targets as well, including the development of novel therapies that activate AMP-activated protein kinase (AMPK) to treat metabolic disorders as part of an ongoing research collaboration and exclusive worldwide license agreement with Eli Lilly and Company. Nimbus has also expanded its pipeline with the addition of new undisclosed therapeutic targets leveraging its novel computational and structure-based drug design approach. "Nimbus made significant progress in 2024 in our discovery and development programs across oncology, immunology, and metabolism, bringing us one step closer to our ultimate goal of delivering transformative medicines to patients," said Jeb Keiper, M.S., M.B.A., Chief Executive Officer of Nimbus. "Peter’s appointment to President of R&D reflects his exceptional leadership and deep expertise in drug discovery and development, and his expanded role will be instrumental in advancing our innovative pipeline. We continue to broaden our drug discovery engine to unlock new difficult-to-drug targets with compelling biology through deep computational expertise, a breadth of internal drug discovery expertise, and key partnerships. We look forward to multiple key milestones in the coming year." Mr. Keiper will provide an overview of the company’s progress, pipeline, and anticipated milestones for 2025 and beyond at the 43 rd Annual J.P. Morgan Healthcare Conference on Monday, January 13, 2025 at 7:30 a.m. PT. About Nimbus Therapeutics Nimbus Therapeutics is a clinical-stage, structure-based drug discovery company developing novel small molecule medicines designed to act against well-validated but difficult-to-drug targets implicated in multiple human diseases. The company advances promising research based on a unique strategy that combines leading-edge computational technologies with a tailored array of machine learning-based predictive modeling approaches. Nimbus' pipeline includes NDI-219216, a Werner syndrome helicase (WRN) inhibitor in development for microsatellite instability high (MSI-H) tumors, as well as a diverse portfolio of preclinical programs focused on cancer, autoimmune conditions, and metabolic diseases. The company is headquartered in Boston, Mass. To learn more about Nimbus, please visit www.nimbustx.com .