Verge Analytics, Inc.

iStock, Robert Way Friday’s deal with fits neatly within AstraZeneca’s business development strategy of upping investments in AI and in China. AstraZeneca put down $110 million upfront to partner with China’s CSPC Pharmaceutical, aiming to leverage AI to develop novel oral therapies for a wide variety of chronic diseases, in a deal announced Friday. Aside from its initial commitment, AstraZeneca also pledged up to $1.62 billion in development milestones, as well as sales-based contingent payments that could reach up to $3.6 billion. All told, Friday’s deal could mean a $5.3 billion windfall for CSPC, plus single-digit royalties on annual net sales of commercial products that emerge from the partnership. AstraZeneca’s payment will grant it access to CSPC’s “AI-driven, dual-engine” drug discovery engine, which looks at how target proteins bind with different compounds, in turn allowing for the identification of small molecule drug candidates with “excellent developability,” as per Friday’s release. The companies did not specify what their priority indications will be, revealing only that the collaboration will include one “small molecule oral therapy for immunological diseases,” which is currently in preclinical development. AstraZeneca will have the right to license candidates under this partnership, which it can exclusively develop and commercialize globally. Friday’s CSPC partnership falls squarely inside AstraZeneca’s business development strategy, in which AI and China are both crucial components. Speaking to BioSpace at the 2025 ASCO meeting earlier this month, Arun Krishna, AstraZeneca’s U.S. oncology head for lung cancer, said the pharma is applying AI at three distinct points in its drug development process: discovery, clinical trials and population screening. According to Krishna, deploying AI across these three critical junctures could not only help the company accelerate and optimize drug development, but could also help AstraZeneca involve itself in the entirety of a patient’s journey. “We primarily want to treat via our drugs. It’s important but not sufficient,” he said. “Our goal is to look at the entirety of the patient pathway. AI is a critical component of that. That’s population-based screening.” To this end, AstraZeneca in recent years has been on an AI dealmaking spree. In September 2023, for instance, the pharma paid $42 million upfront and promised $840 million in milestones to partner with Verge Genomics. The multi-target, four-year deal will leverage the biotech’s drug discovery AI for rare diseases. More recently, in April, AstraZeneca joined hands with Tempus to create a cancer-focused AI model that can help collect biological and clinical insights, which can then be used to guide drug development. Friday’s CSPC collaboration also cements AstraZeneca as a solid investor in China—despite mounting political pressure to move away from the Asian giant. In February, the pharma snapped up FibroGen’s China business for $160 million, which it followed up a month later with a potential $10 billion -plus investment in the country.

关注并星标CPHI制药在线       阿斯利康已在人工智能领域的合作中投入了数亿美元的资金，其目标不仅是加速开发药物来治疗已确诊的癌症，还包括开发能够比现有方法更早发现癌症的诊断技术。1. 肿瘤学负责人谈阿斯利康AI药物开发三阶段过去几年，阿斯利康已向一系列人工智能合作项目投入了超过10亿美元的生物医学资金（Biobucks）。在芝加哥ASCO 2025（美国临床肿瘤学会年会）会议期间，阿斯利康美国肺癌肿瘤学负责人阿伦·克里希纳（Arun Krishna）接受采访并分享了这些资金的用途。他说："人工智能是一个非常广泛的话题，我按生命周期来分解一下。"据克里希纳介绍，人工智能药物开发分为三个阶段。最受关注的阶段是药物发现阶段，英矽智能（Insilico）和Generate:Biomedicines等公司已在此领域占据一席之地。在这个阶段，相比在湿实验室中需要数月或数年的过程，预测性人工智能有望更快地识别出潜在有用的分子。现在，这个过程可以在30天或更短的时间内完成。第二个阶段是临床试验阶段。我们能否使用现有的预测性人工智能来标记可能符合条件的患者？这不仅在于分析现有患者数据并选择可能符合条件的患者（人类也能很好地完成这一点），而是预测哪些患者可能从新候选药物中获益最大。例如，患者可能尚未出现直接符合参加临床试验条件的症状或生物标志物，但他们未来可能会出现这些症状或生物标志物，而人工智能可能很适合根据其档案中的其他线索做出判断。人工智能还可能识别出已经符合条件但尚未进行适当检查的患者。克里希纳说："即使我们不知道某个具体的生物标志物，如果患者符合某种特征，你就应该检测该相关生物标志物。人工智能可以为你填补这一空白。"监管机构、临床试验组织者和学术中心仍在积极讨论这一领域。克里希纳预测，基于人工智能的临床试验患者招募可能在1-2年内成为药物测试领域的一个活跃领域。第三个阶段最为不同寻常：人群筛查。据克里希纳介绍，阿斯利康的下一步不仅是治疗癌症，还将涉足诊断领域。克里希纳举了一个例子：偶发性肺结节（IPN）在人群中经常会出现。大多数情况下，这些结节是良性的，不会引发任何问题，但在5%-6%的患者中，它们会癌变。放射科医生在阅读CT扫描图像时有可能漏诊这些结节。克里希纳说，阿斯利康已将其预测性人工智能算法与人类放射科医生进行了并列比较。人工智能能够在放射科医生认为没有问题的病例中发现癌性结节。需要明确的是，过去几十年里，偶发性肺结节的检测率有所上升，这主要是由于医生开具的扫描检查数量增加，而且绝大多数结节是良性的。但正如克里希纳所说，阿斯利康希望在患者疾病的整个过程中都发挥作用。人工智能是其中的关键组成部分，即在人群癌症筛查中。2. 阿斯利康广泛积极的AI合作为了充分挖掘人工智能在各个方面的潜力，阿斯利康一直在招揽合作伙伴。今年4月，阿斯利康与Tempus AI和Pathos达成三方合作，投入2亿美元用于数据许可，以构建多模态人工智能肿瘤学模型，这对建立数据基础至关重要。在此之前，阿斯利康还达成了其他几项人工智能的合作。2023年12月，阿斯利康向人工智能开发公司Absci承诺提供高达2.47亿美元的资金，目标是在某个未披露的肿瘤学领域寻找新靶点。同样在2023年底，阿斯利康向Verge Genomics预先支付了4200万美元，并承诺在达到里程碑时支付8.4亿美元，以利用这家生物技术公司的人工智能平台在神经退行性和神经肌肉疾病领域寻找新靶点。阿斯利康还与BenevolentAI和人工智能免疫肿瘤学公司Immunai达成了多项合作。在人群筛查方面，阿斯利康正在与马里兰大学癌症医院的格林鲍姆癌症中心（Greenbaum Cancer Center）合作。克里希纳说："我们希望大规模推广这一技术，并在18个月内完成5万次CT扫描。"结语阿斯利康在人工智能领域的布局，展现出其对癌症诊疗全链条革新的决心。通过多阶段药物开发与广泛合作，它正以AI为引擎，加速突破传统医疗局限，从早期筛查到精准治疗全面发力，有望为癌症诊疗模式带来颠覆性变革，开启智能医疗的全新篇章。参考来源：       AstraZeneca Racks Up AI Partners, Seeking To Stop Cancer at All Stages, June 2, 2025, BioSpaceEND【企业推荐】来源：CPHI制药在线声明：本文仅代表作者观点，并不代表制药在线立场。本网站内容仅出于传递更多信息之目的。如需转载，请务必注明文章来源和作者。投稿邮箱：Kelly.Xiao@imsinoexpo.com▼更多制药资讯，请关注CPHI制药在线▼点击阅读原文，进入智药研习社~

iStock, SvetaZi AstraZeneca has put hundreds of millions of dollars into AI deals, with an eye toward not just accelerating the development of drugs that treat cancer after it appears but also in creating diagnostics that can catch cancer earlier than current methods allow. AstraZeneca has put more than $1 billion in biobucks into a series of artificial intelligence partnerships over the past few years. In the basement of Chicago’s McCormick Place, in the spooky, cubicle-like temporary meeting rooms that ASCO 2025 sets up for meetings, I asked the company’s head of U.S. oncology Arun Krishna what exactly that money was paying for. He leaned back in his squishy office chair before answering. “AI is such a huge topic,” he said. “Let me break it down into a life cycle.” According to Krishna, there are three stages of AI drug development. The buzziest one, where companies like Insilico and Generate:Biomedicines have staked their claims, is in discovery. That’s where predictive AI can, hopefully, identify potentially useful molecules much faster than the months or years the process would take in wet labs. “Now it can be 30 days or less,” Krishna said. The second stage is in clinical trials. “Can we use predictive AI that we have to flag patients who might be eligible?” Krishna asked. According to Krishna, this is not just a matter of analyzing existing patient data and selecting people who might be eligible, something a human can do just fine, but a matter of predicting who might benefit the most from a new drug candidate. A patient, for example, might not have developed symptoms or biomarkers that directly qualify them for a clinical trial, but they could develop those down the line, and AI might be well suited to make that determination based on other clues in their profile. It may also be able to identify patients who are already eligible but just haven’t gotten the proper tests. “Even if we don’t know a biomarker,” Krishna said, “if [a patient] fits a profile, you should test for that inclusive biomarker. AI fills in the gap for you.” This is still an active area of discussion for regulators, clinical trial organizers and academic centers. But Krishna forecasts that AI-based clinical trial patient enrollment could be an active part of the drug testing landscape “within 1-2 years.” The third stage is the most unusual: population screening. According to Krishna, the next step for AstraZeneca is not just in treating cancer but in getting involved at the diagnostic level. Krishna gave an example: people often develop incidental pulmonary nodules (IPNs). Most of the time they are benign and lead to nothing, but in 5–6% of patients they become cancerous. They can also be missed by radiologists reading a CT scan. Krishna said that AstraZeneca has done side-by-side comparisons of its predictive AI algorithm versus a human radiologist. The AI can spy a cancerous nodule where a radiologist would say there is nothing. To be clear, there’s been a rise over the last few decades in the detection of IPNs, mostly driven by increased numbers of scans being ordered, and the vast majority are benign, but as Krishna put it, the company wants to be involved in the entire arc of a patient’s illness. “It is unusual,” Krishna said. “We primarily want to treat via our drugs. It’s important but not sufficient. Our goal is to look at the entirety of the patient pathway. AI is a critical component of that. That’s population-based screening.” Partnering For Success To take advantage of AI’s capabilities across all board, AstraZeneca has been scooping up collaborators. “We know that we’re not an AI company but that we have to partner,” Krishna said. In April, AstraZeneca entered into a three-way partnership with Tempus AI and Pathos , with $200 million in data licensing, to build a multimodal AI oncology model. “This is about building those data foundations,” Krishna said. That followed several other AI-based deals made by AstraZeneca. In December 2023, the pharma committed up to $247 million to AI development company Absci with the goal of finding new targets in unspecified areas of oncology. Also in late 2023, AstraZeneca gave Verge Genomics $42 million up front, with $840 million in milestone payments on the line, to use the biotech’s AI platform to find new targets in neurodegenerative and neuromuscular diseases. AstraZeneca has also struck multiple collaborations with BenevolentAI and AI immuno-oncology firm Immunai. When it comes to population screening, the pharma is working with Greenbaum Cancer Center at the University of Maryland Cancer Hospitals. “We’re looking to roll this out at scale. We’re looking to get up to 50,000 CT scans within 18 months,” Krishna said. It’s not difficult to understand the potential role AstraZeneca sees AI playing in drug development. The company has put more than $1 billion down on it, eagerly sent Krishna to talk with me about it. And when Krishna speaks, it erases all doubt: “We believe that AI is the future. It will really enable every aspect of that patient journey and ultimately serve our purpose of eliminating cancer. We want to play an integral role in that.”