Nature 子刊｜从算法开发到2期临床，详解全球首款“AI药物”研发历程

Swipe left for English

• 近日发表于Nature Biotechnology的一项研究，首次阐述全球首款人工智能（AI）药物INS018_055的研发历程，涵盖从AI算法开发到2期临床试验的全流程； 

• 关于该项目的原始实验数据，可通过访问英矽智能的数据中心(https://insilico.com/repository/nbt-ins018-055-tnik)获取，包括13项临床前实验和3项临床试验的部分数据； 

• 英矽智能在新一代大语言模型ChatGPT-4 Turbo和自有大语言模型（LLM）基础上开发了PaperGPT论文解读引擎(https://papers.insilicogpt.com/)，通过语言对话功能提供与论文相关问题的专业解答。 

现代医学发展至今，世界上仍有数千种疾病面临“无药可医”甚至“无药可用”的困境。而传统药物发现耗时漫长成本高昂，且伴随着极高的失败率，超过90%的候选药物在关键的临床验证阶段折戟。人工智能（AI）的出现为流程优化与效率提升带来了希望，但真正开启AI制药时代，还需要真实世界的有力验证，包括临床前实验验证和临床阶段验证。 

近日，临床阶段生成式AI驱动的生物科技公司英矽智能发表于Nature Biotechnology上的一项研究提供了这一例证。研究全面阐述了其首款由生成式AI发现和设计的潜在“全球首创”（first-in-class）TNIK抑制剂从人工智能算法开发到2期临床试验的研发历程，并首次披露了该候选药物在临床前实验和临床试验中的数据和表现。该研究突出了AI驱动的药物发现方法带来的降本增效优势，并强调了生成式AI技术在推动行业变革方面的巨大潜力。 

Nature Biotechnology发文截图，
点击文末【阅读原文】即可访问

英矽智能创始人兼首席执行官Alex Zhavoronkov博士表示，“我认为分享INS018_055的进展对药物发现领域具有重大意义。这不仅是英矽智能端到端药物发现平台 Pharma.AI 的概念验证，还开创了生成式AI加速发现创新药物的先例。从这篇论文中，我们看到生成式AI简化早期药物发现的潜力，该解决方案的扩展应用有望解决行业面临的成本和效率挑战，高效提供创新疗法尽快惠及全球病患。” 

作为新药研发的第一步，英矽智能选择以与衰老密切相关的纤维化作为研究重心，采用组织纤维化相关的组学和临床数据集，对Pharma.AI平台下属的靶点发现引擎PandaOmics进行训练。以此为基础，PandaOmics平台通过深度特征合成、因果关系推断和全新通路重建等过程提名潜在靶点列表。此后，PandaOmics中的自然语言处理（NLP）模型通过分析了涵盖专利、出版物、研发基金、临床试验等文本数据的数百万个文本文件，进一步评估潜在靶点的新颖性以及与疾病的关联性，最终确定TNIK为最有潜力的抗纤维化靶点。值得注意的是，历史研究曾揭示TNIK与多种纤维化驱动生物通路的间接关联，但从未提出将其作为特发性肺纤维化（IPF）治疗靶点。 

确定TNIK靶点后，英矽智能研发团队利用Pharma.AI下属的生成化学引擎Chemistry42，根据基于结构的药物设计（SBDD）策略生成具有所需特性的创新分子结构，旨在得到安全、特异性、高效的TNIK抑制剂。Chemistry42 结合了 40 多种生成化学算法和超过500个预训练的奖励模型，支持新颖化合物从头生成，能根据专家反馈进行虚拟筛选并优化生成结果。经过多次迭代筛选，团队发现了IC50值达到纳摩尔级别的潜力苗头化合物，并在针对溶解度、ADME安全性、毒性进行优化的同时保留其对 TNIK 的显著亲和力，最终获得了候选分子 INS018_055，共合成并测试了不到80个分子。 

在随后的临床前研究中，INS018_055 在体内和体外试验中均显示出对 IPF 的显著疗效，并在多个细胞系和多个物种的药代动力学和安全性研究中显示出良好的结果。此外，INS018_055 还表现出泛纤维化抑制功能，在另外两种动物模型中减轻了皮肤和肾脏纤维化。基于这些研究，INS018_055 于 2021 年 2 月被提名为临床前候选化合物。此时距离TNIK被PandaOmics提名为潜在IPF治疗靶点，仅仅过去了18个月。 

在人体临床研究中，INS018_055也交出了出色的答卷。2021 年 11 月，在获得 PCC 提名 9 个月后，INS018_055 在澳大利亚的首次人体微剂量试验中完成首批健康受试者给药。该项人体微剂量试验结果超出预期，展现了候选药物良好的药代动力学和安全性特征， 不仅完成了AI制药临床概念验证，还为后续临床试验奠定了基础。在新西兰和中国进行的 I 期试验中，INS018_055 分别在 78 名和 48 名健康受试者中进行了测试，完成了单次剂量递增 （SAD）和多次剂量递增（MAD）队列研究。国际多中心1期临床试验得出了一致的结果，表明 INS018_055 具有良好的安全性、耐受性和药代动力学 (PK) 特征，支持后续2期临床试验开展。 

英矽智能联合首席执行官兼首席科学官任峰博士表示，“结合人工智能方法与人类智能经验，我们成功提名了INS018_055这款具有全球首创潜力的新颖抗纤维化抑制剂，同时在早期药物发现过程中大幅降低了时间和成本投入。基于积极的临床前和已有临床数据，我们期待INS018_055在2期临床研究中有良好的表现，为患者提供新的治疗选择，同时为AI制药行业带来更坚实的例证。” 

截至文章发表，使用INS018_055治疗IPF的两项随机、双盲、安慰剂对照2a期临床试验正在中美两地同步开展中，旨在评估候选药物的安全性、耐受性、药代动力学特征，并评估其针对IPF患者肺功能的初步疗效。INS018_055的持续进展有望为全球500万罹患这种致命疾病的患者带来希望。 

英矽智能的药物发现流程由其经验证、可商业化的自研人工智能药物研发平台Pharma.AI驱动，该平台横跨生物、化学和临床医学领域，作为先进的生成式人工智能工具，为生物制药行业提供研发助力。在 Pharma.AI 的支持下，英矽智能正在纤维化、肿瘤、免疫、老龄化相关疾病等多个领域为医疗保健带来突破。自2021年来，英矽智能已搭建涵盖30余条管线的多元化疗法组合，提名18款临床前候选化合物，并将其中6个领先项目推进到临床阶段。 

参考资料

[1] Ren, F., et al. A small-molecule TNIK inhibitor targets fibrosis in preclinical and clinical models. Nat Biotechnol (2024). https://doi.org/10.1038/s41587-024-02143-0

关于英矽智能

英矽智能是一家由生成式人工智能驱动的临床阶段生物医药科技公司，通过下一代人工智能系统连接生物学、化学和临床试验分析，利用深度生成模型、强化学习、转换模型等现代机器学习技术，构建强大且高效的人工智能药物研发平台，识别全新靶点并生成具有特定属性分子结构的候选药物。英矽智能聚焦癌症、纤维化、免疫、中枢神经系统疾病、衰老相关疾病等未被满足医疗需求领域，推进并加速创新药物研发。

更多信息，请访问网站
www.insilico.com

商务合作，请联系 [e-mail]

媒体垂询，请联系 [e-mail]

右滑查看中文版

• A study published in Nature Biotechnology presents the entire journey of INS018_055, from AI algorithms to Phase II clinical trials for the first time.
  

• Raw data from 13 preclinical experiments and 3 clinical trials referenced in this study can be accessed by visiting Insilico’s data room(https://insilico.com/repository/nbt-ins018-055-tnik).

• Insilico developed a PaperGPT system based on ChatGPT-4 Turbo and internal LLM that provides answers related to the paper via chat functionality (https://papers.insilicogpt.com/). 

There are thousands of diseases worldwide with no cure or available treatments. Traditional drug discovery and development takes decades and billions of dollars and more than 90% of these drugs fail in clinical trials. The emergence of artificial intelligence (AI) holds promise for streamlining and improving the entire process. However, ushering in a new era of AI-driven drug discovery requires costly and lengthy validation in preclinical cell, tissue, and animal models and human clinical trials. 

Now, that preclinical and part of that clinical validation was published in a new study in Nature Biotechnology. In this paper, Insilico Medicine and collaborators present the journey of its lead therapeutic program with an AI-discovered target and novel molecule generated from AI algorithms to Phase II clinical trials. For the first time, the paper discloses the raw experimental data and the preclinical and clinical evaluation of the potentially first-in-class TNIK inhibitor discovered and designed through generative AI. The study underscores the benefits of AI-led drug discovery methodology to provide efficiency and speed to drug discovery and highlights the promising potential of generative AI technologies for transforming the industry. 

“When our first paper in the generative AI for generation of novel molecules was published in 2016, followed by many follow-up papers, the drug discovery community was very skeptical. Even after several validation experiments and launch of our AI software platform that is now used by many biopharma companies, many questions remained. Based on the research data, especially those from the live clinical program. To date, I have not seen anything close from any other company in our field,” said Alex Zhavoronkov, PhD, founder and CEO of Insilico Medicine. “From my perspective, the progress of INS018_055 has significant implications for the drug discovery field. It not only serves as a proof-of-concept for Pharma.AI, our end-to-end AI-driven drug discovery platform, but sets a precedent for the potential of generative AI to accelerate drug discovery. Using the publication as a guide, one can extrapolate how generative AI drug discovery tools may streamline early discovery efforts. We anticipate that the expanded application of this platform will address challenges facing industry R&D, including cost and efficiency, and accelerate the delivery of innovative therapies to patients.”

Insilico initiated the research by focusing on fibrosis, a biological process closely associated with aging. The group first trained PandaOmics, the target identification engine of Insilico’s proprietary AI platform Pharma.AI, on the collection of omics and clinical datasets related to tissue fibrosis. Next, PandaOmics proposed a potential target list using deep feature synthesis, causality inference, and de novo pathway reconstruction. After that, the natural language processing (NLP) models of PandaOmics analyzed millions of text files, including patents, publications, grants, and clinical trial databases to further assess the novelty and disease association. TNIK was identified as the most promising anti-fibrosis target. Notably, TNIK had been indirectly linked to multiple fibrosis-driven pathways in previous research but was never pursued as a potential target for IPF. In a separate paper, Insilico scientists demonstrated that TNIK may be implicated in multiple hallmarks of aging. 

After selecting TNIK as a primary target, Insilico scientists utilize Chemistry42, the Company’s generative chemistry engine, to generate novel molecular structures with the desired properties using the structure-based drug design (SBDD) workflow. Chemistry42 combines over 40 generative chemistry algorithms and over 500 pre-trained reward models for de novo compound generation, and can optimize both generation and virtual screening based on expert human feedback. Following multiple iterative screens, one promising hit candidate demonstrated activity with nanomolar IC50 values. The group further optimized the compound to increase solubility, promote a good ADME safety profile, and mitigate unwanted toxicity while retaining its remarkable affinity for TNIK, which ultimately produced the lead molecule INS018_055, with less than 80 molecules synthesized and tested.

In subsequent preclinical studies, INS018_055 demonstrated significant efficacy in vitro and in vivo studies for IPF and showed promising results in pharmacokinetic and safety studies across multiple cell lines and multiple species. Furthermore, INS018_055 showed pan-fibrotic inhibitory function, attenuating skin and kidney fibrosis in two additional animal models. Based on these studies, INS018_055 achieved preclinical candidate nomination in February 2021, in less than 18 months following PandaOmics’ proposal of TNIK as a potentially novel target for IPF in 2019.

INS018_055 has exhibited excellent performance in clinical trials to date. In November 2021, 9 months after PCC nomination, the first healthy volunteers were dosed in a first-in-human (FIH) microdose trial of INS018_055 in Australia. This microdose trial exceeded expectations, delivering a favorable pharmacokinetic and safety profile that successfully demonstrated this clinical proof-of-concept and set the stage for the next step of clinical testing. In Phase I trials carried out in New Zealand and China, INS018_055 was tested in 78 and 48 healthy subjects, divided into cohorts focusing on a single ascending dose (SAD) study and multiple ascending dose (MAD) study. The international multi-site Phase I studies yielded consistent results, demonstrating favorable safety, tolerability, and pharmacokinetics (PK) profiles of INS018_055, and supporting the initiation of the Phase II studies.

“Combining AI methods with human intelligence, we have successfully nominated INS018_055, a potentially first-in-class antifibrotic inhibitor, with significant reductions in time and costs”, said Feng Ren, PhD, co-CEO and Chief Scientific Officer of Insilico Medicine. “Encouraged by positive preclinical and available clinical data, we look forward to favorable performance of INS018_055 in Phase 2 clinical trials, which would provide innovative options for patients while bringing more solid evidence for the AI-driven drug discovery industry.” 

At the time of this publication, two Phase 2a clinical trials of INS018_055 for the treatment of IPF are being conducted in parallel in the United States and China. The studies are randomized, double-blind, placebo-controlled trials designed to evaluate the safety, tolerability and pharmacokinetics of the lead drug. In addition, the trials will assess the preliminary efficacy of INS018_055 on lung function in IPF patients. As this drug continues to advance, it drives hope for the roughly five million people worldwide suffering from this deadly disease.

Insilico’s drug discovery efforts are driven by its validated and commercially viable AI drug discovery platform, Pharma.AI, which works across biology, chemistry, and clinical medicine, providing the biotechnology and the pharmaceutical industry with advanced generative AI tools to accelerate their internal research and development. Powered by Pharma.AI, Insilico is delivering breakthroughs for healthcare in multiple disease areas, including fibrosis, cancer, immunology and aging-related disease. Since 2021, Insilico has nominated 18 preclinical candidates in its comprehensive portfolio of over 30 assets and has advanced six pipelines to the clinical stage.

Reference

[1] Ren, F., et al. A small-molecule TNIK inhibitor targets fibrosis in preclinical and clinical models. Nat Biotechnol (2024). https://doi.org/10.1038/s41587-024-02143-0

About Insilico Medicine

Insilico Medicine, a global clinical stage biotechnology company powered by generative AI, is connecting biology, chemistry, and clinical trials analysis using next-generation AI systems. The company has developed AI platforms that utilize deep generative models, reinforcement learning, transformers, and other modern machine learning techniques for novel target discovery and the generation of novel molecular structures with desired properties. Insilico Medicine is developing breakthrough solutions to discover and develop innovative drugs for cancer, fibrosis, immunity, central nervous system diseases, infectious diseases, autoimmune diseases, and aging-related diseases. www.insilico.com