您正在使用IE低版浏览器，为了您的雷峰网账号安全和更好的产品体验，强烈建议使用更快更安全的浏览器
此为临时链接，仅用于文章预览，将在时失效
0
导语：“相较于复杂的生物学预测，我们目前更看重AI在海量筛选方面的应用价值。”
作者丨周蕾
编辑丨岑峰
【编者按】被誉为"医药界诺贝尔奖"的美国盖伦奖（Prix Galien USA），于10月29日公布了2026年度获奖名单。
值得注意的是，晶泰科技孵化的创新企业已连续两年入围该奖项提名。去年，希格生科凭借全球首款弥漫性胃癌靶向药SIGX1094获得"最佳生物技术产品"提名；今年，莱芒生物与默达生物双双入围“最佳初创公司”提名——前者在晶泰研发助力下，以千分之一常规剂量的代谢增强型CAR-T疗法获得提名，后者则与晶泰合作研发了全球首款口服LDH抑制剂MP-5342，瞄准全球超千亿美元的自免市场。
在奖项揭晓之后，雷峰网AI科技评论先后与希格生科、莱芒生物、默达生物进行了深度对话，以下是默达生物篇。
2021年春天，康奈尔医学院的许柯收到两份邀请。两位在免疫领域颇有声望的美国教授，各自创办了一家免疫治疗公司，都邀他作为联合创始人加入。
待遇优厚，方向明确，一切准备就绪——许柯拒绝了。
“我为什么不去做我自己的东西？”
彼时，在康奈尔和纪念斯隆-凯特琳癌症中心的联合培养项目中，许柯已经花了近八年时间研究免疫与代谢的交叉地带。他手里有一组数据，关于免疫细胞在发动攻击前如何改变自己的代谢基础。这组数据的潜力，在他看来，比任何现成公司的offer都大。
默达生物的CTO鲜于安今，当时还在美国最大的病人数据公司PatientsLikeMe做数据科学总监，薪资极高。但许柯说服鲜于安今时，并没有花多少力气，“我的数据，他看一看，就知道这是多好的东西。”
两人一拍即合，决定回国创业。晶泰科技是他们最早的投资人之一，两人回国后就直接落脚在晶泰的办公楼里，成为“楼上楼下”的创新伙伴。
许柯瞄准的免疫代谢赛道，即便在国外，也仍处于持续探索阶段。代谢网络中的酶种类繁多，但针对免疫细胞代谢开发新药，仍面临选择性与安全性等难题。
代谢太基础了，干预它意味着牵一发而动全身，安全性风险难以预测。但许柯从2013年开始就在这个领域深耕，他观察到一个关键现象：部分参与自免疾病的免疫细胞，在异常激活过程中会发生 “代谢重编程”——增加葡萄糖利用，以支持细胞功能。
“坏细胞干坏事之前，得先吃饱饭。”许柯这样比喻。默达生物希望通过干预特定代谢环节，调节异常免疫反应，探索自身免疫性疾病及代谢性疾病的新治疗方式。
默达研发的口服小分子LDHA抑制剂MP-5342，相当于“饿着”免疫系统的坏细胞，通过抑制糖酵解通路，降低免疫细胞的代谢活性，让免疫反应回归稳态，而不是像传统免疫抑制剂那样把免疫系统猛压到谷底。
这种“饥饿疗法”带来了超过600倍的安全窗口，远超行业常规安全阈值（通常为>40倍）。正常细胞平时并不依赖这条通路，只有处于应激状态的病理细胞才高度需要它。相比之下，年销售额曾接近200亿美元的修美乐（阿达木单抗），以及当下的JAK抑制剂，都带有黑框警告——严重感染风险，甚至心血管风险。
默达没有从病人基数最大的银屑病或特应性皮炎入手，而是首选炎症性肠病（IBD）。这种疾病的现有药物普遍面临临床缓解率不足40%的困境，严重患者一天可能需要上20次厕所，甚至出现肠穿孔等并发症。此外，原发性胆汁性胆管炎、多发性硬化症等仍有未满足临床需求的领域也在拓展清单上。
基于同样的免疫代谢底层生物学机制，默达生物发现了一系列代谢蛋白酶的全新靶点，这些靶点处在代谢网络的关键节点，具有拓展至多种不同适应症的潜力。除了自免疾病，默达还在推进针对原发性高草酸尿症的全球首创（First-in-Class）口服小分子抑制剂 META-001-PH——一种儿童期多发的罕见遗传性代谢病，患儿体内草酸不断累积，多数在青春期前进入肾衰竭。许柯告诉我们，这种病国内尚无特效药。
在研发路径上，默达选择了与传统方法“相反方向”的方式。传统药物研发多是从细胞或动物模型往人身上推，默达则反过来，从IBD病人的多组学数据出发，寻找患者群体中共同上调的代谢通路，再回到动物身上做基因敲除验证，评估这个通路是否参与疾病的发生发展，并进一步针对它开发抑制剂。
AI正好可以帮助他们将药物发现从“大海捞针”转向“精准导航”。晶泰的AI+机器人平台，可以结合项目的研发目标，负责从海量的虚拟分子库中，快速识别可能与特定蛋白靶点结合的潜在候选药物，帮助选择更有潜力的分子进入合成与实验验证；默达自建的AI平台则负责分析临床样本的基因组、代谢组和蛋白组数据。
但许柯对AI的态度很清醒：生物不是物理，不是0和1，中间有巨大的灰色地带；同一个protocol，不同的人在不同的城市，做出来的结果可能都不一样。因此，在他看来，相较于复杂的生物学预测，默达现阶段更看重AI在海量筛选方面的应用价值。
按照计划，MP-5342将于2027年进入健康人临床I期，2028年进入IIa期病人临床试验。默达生物也即将完成新一轮融资，以支持后续临床开发。
在AI制药经历高峰与低谷的周期里，默达的笃定或许来自一种更底层的坚持——把近十年的学术积累，变成一款加速步入临床、尽早惠及病患的分子。
以下是默达生物创始人兼CEO许柯与雷峰网(公众号：雷峰网)AI科技评论的对话节选，有不改变原意的编辑：
01
创业抉择：回国还是留下？
2021年，许柯在康奈尔医学院和纪念斯隆-凯特琳癌症中心完成博士联合培养，手握关于免疫代谢交叉领域的关键数据。当两位美国教授向他抛出免疫治疗创业公司的联创邀请时，他选择了一条更艰难的路——用免疫代谢重新定义自免疾病的治疗可能。
▎AI科技评论：当时美国那边给了很好的待遇，也有教授邀请你做联创，为什么还是选择回国创业？
许柯：当时离开的时候，我们学校有好几个特别牛的教授，刚好在我这个免疫领域，他们自己也做了创业公司，有两位教授都邀请我作为联合创始人加入，去做免疫相关治疗药物的开发。那是非常好的机会，不用我自己辛苦，他们已经把摊子做起来了。
但那时候我觉得，我自己做的东西更有意思——与其选择做一个联创去做别人的东西，我为什么不去做我自己的东西？那时候我已经有很坚定的要创业的想法了。
▎AI科技评论：鲜于安今博士当时是怎么被你说服的？
许柯：这个项目一直都是我之前在研究的东西，我就邀请鲜于安今博士加入。
他当时已经工作了，在美国最大的一个病人医疗数据平台PatientsLikeMe，做到director级别，那边给的工资待遇非常高。但说服他没有花很大力气，大家都是搞科研的，我不需要用情怀或者别的什么去说服他，我就说：这是我的数据，你看一看。我不用多讲，你就知道这是多好的东西、潜力有多大。作为科学研究者，用数据说话是最能说服人的。
▎AI科技评论：回国为什么落在深圳？
许柯：其实当时对落哪个城市没有特别明确的偏向，上海也好，苏州也好，深圳也好，都可以。刚好主要投资人和业务合作伙伴晶泰科技在这边，未来还有很多合作机会，就落在深圳了。落地的时候，就是直接在晶泰内部开始办公的。
▎AI科技评论：和晶泰最早是怎么认识的？
许柯：最早是一家叫IMO Ventures的早期科技基金，他们看到我们发表的文章，觉得我们做的免疫代谢方向特别好。刚好IMO也是晶泰的投资者，觉得我们既然做小分子药物，和晶泰在AI制药这块应该能互补，就介绍我们和书豪总认识，回国前我们就把局组起来了。
▎AI科技评论：从科学家转向企业家，适应起来困难吗？
许柯：对我来说还好。从科学家转向企业家，需要在科研之外，逐步熟悉团队管理、融资和产品开发，确实有不少新课题。我当时也处在探索职业方向的阶段，愿意尝试不同的工作方式，所以比较容易进入边做边学、不断调整的状态。
▎AI科技评论：做CEO以来，有没有一些观念上的变化？
许柯：我一直想的就是怎么把蛋糕做大，希望通过长期研发与团队合作创造价值，也重视与投资人的长期共识。这是我创业里比较强烈的一个心得。
▎AI科技评论：你对默达的定位，是做平台还是做管线？
许柯：我自己的认知是，我们就是一个管线公司——所谓的那些平台，都是服务于我们去发展管线的。平台属性有，但我的定位就是管线，主要以发展产品为主。
▎AI科技评论：未来12个月，默达最想要达成的一个目标是什么？
许柯：就是推进MP-5342进入临床，以及完成新一轮融资，支持后续临床开发——药物研发需要持续投入。
02
技术路径：让坏细胞“吃不饱饭”
默达的核心技术建立在许柯自2013年起积累的免疫代谢研究之上。与免疫抑制或细胞清除的传统思路不同，默达提出了一种“代谢检查点”机制：让攻击自身组织的免疫细胞在“黑化”之前，“吃不饱饭”，从而在不造成免疫缺陷的前提下恢复免疫稳态。
▎AI科技评论：能通俗解释一下默达的技术原理吗？
许柯：自身免疫疾病发病诱因很多，但免疫失调是共同的核心环节：IBD主要是针对肠道菌群的异常T细胞（及固有免疫）反应驱动，而红斑狼疮则以T细胞辅助下的B细胞过度活化、产生致病性自身抗体为突出特征，都是免疫功能失调过激、不识别敌我。
这些免疫细胞有一个很大的共同特征：它在变成导致疾病发展的坏细胞之前，会有一个代谢重编程的过程——它要变成坏人，首先要改变自己的代谢基础。你可以想象，一个坏人干坏事之前，得先吃饱饭，没有力气怎么干坏事？打仗之前，粮草先行。这个代谢尤其体现在葡萄糖代谢上，坏细胞会大量吸收葡萄糖获取能量。
我们的策略就是：你既然必须先吃饱饭才能干坏事，那我就不让你吃饱饭——通过代谢酶抑制剂，把它代谢能量通路的活性压下来，类似于一种“饥饿疗法”。
▎AI科技评论：这种“断粮”会不会把正常免疫细胞也饿死？
许柯：不会。免疫系统一直处于动态平衡。免疫过激导致自免，但被抑制得太厉害就类似免疫缺陷，很容易受到细菌、病毒、真菌感染，感染风险非常严重。它一定是平衡状态，不能过激也不能过低。
我们的治疗方法和现有药不一样。像阿达木单抗（修美乐）、JAK抑制剂这些，是猛地把免疫能力全部抑制掉，所以有黑框警告，严重感染风险，还有心血管风险。我们是让免疫状态回归稳态、平衡状态，不需要压得很低。
▎AI科技评论：这个通路的安全性顾虑，你会怎么回应？
许柯：这也是投资人问得最多的，“你干扰葡萄糖代谢，风险会不会很大？”其实我们做了大量实验证明，这个通路是应激状态下的通路，平时细胞不怎么使用它——平时细胞就是有氧呼吸，葡萄糖进入线粒体产生大量能量；而糖酵解通路是在无氧状态、剧烈运动时才需要的，正常稳态下不太需要。所以抑制它对正常细胞的功能和生存来说影响相对有限，但对处于病理状态的免疫细胞则展现出良好的抑制作用。
▎AI科技评论：自免方向还有一类路径是CAR-T，你们的路径有什么不同？
许柯：部分用于自免疾病研究的CAR-T疗法，通过清除表达特定抗原的B细胞等细胞群，尝试重置免疫系统。这类治疗及其预处理可能带来感染等风险，需要严密管理。我们的方向是通过代谢调节，不是清除。自免疾病的免疫细胞本身还是正常细胞，只是识别错了目标。我们让它回到平衡就好。
▎AI科技评论：为什么首选IBD？银屑病或皮炎的病人应该更多一些。
许柯：做药一定要和临床需求相关——不能只看病人的数量，更要看现有疗法的效果和未被满足的临床需求。银屑病和特应性皮炎的一线疗法已经较为成熟且有效，临床需求相对小。所以，我们更希望将有限的研发资源集中在临床需求更为迫切的疾病领域。
IBD（炎症性肠病）现有药物的完全缓解率不到40%，临床需求非常大。它主要分两种：溃疡性结肠炎和克罗恩病，稍微严重一点的病人一天可能需要上20次厕所，随时随地找厕所，不敢外出，一直拉肚子，严重的会肠穿孔、消化道出血。成因就是自身免疫——你的免疫系统攻击你的肠道，造成长期慢性炎症。
而且，IBD的病人需要反复调整治疗方案，因为现有已上市批准的药，临床完全缓解率仍然有限，部分药只针对某一个患者亚群，难以满足所有病人的治疗需求。所以非常需要一个既安全又有效的药。
▎AI科技评论：其他适应症是怎么筛掉的？
许柯：我们在动物模型上做过特应性皮炎、多发性硬化症、原发性胆汁性胆管炎、银屑病、类风湿性关节炎，效果都非常好。多发性硬化症是免疫系统攻击神经元，导致逐渐全身瘫痪，机制跟我们很相近，无非攻击的器官不一样。自免疾病现在有100多种，每个器官都有好几种。这里面临床需求最大的是IBD；还有原发性胆汁性胆管炎、多发性硬化症，也仍有未满足的治疗需求。
▎AI科技评论：同一个靶点还做了罕见病管线？
许柯：对，还有一条管线是相同靶点、不同分子，做原发性高草酸尿症，儿童多发的、非常致命的罕见病。这些小孩先天基因缺陷，肝脏产生大量草酸，草酸和钙离子结合变成草酸钙，就是肾结石，从出生那一刻就在累积，几个月大的婴儿就会有肾结石。目前国内没有特效药，治疗选择仍然有限，部分患者需要依赖水化、药物干预及结石管理等措施，也存在较大的未满足临床需求。
对于病情严重的患者，可能在儿童或者青少年时期就发展为肾衰竭，需要接受器官移植。由于部分类型的病源在于肝脏的代谢异常，因此部分患者需要接受肝肾双移植，单纯换肾解决不了问题。这个病跟免疫看起来不相关，但刚好我们这个代谢酶可以抑制草酸生成——它不仅是催化糖代谢的节点，还是催化草酸生成的节点，是一个多功能的代谢酶。所以同一个靶点可以做两个方向的适应症。
▎AI科技评论：代谢酶这个方向为什么做的人少？
许柯：代谢网络里还有很多值得探索的靶点，尤其是针对免疫细胞的药物开发，很多潜力大家还没意识到。不是技术上不允许，更多是大家对于代谢系统偏保守，觉得代谢很重要、干预了安全性有风险；而且免疫代谢这个交叉学科还处于早期，注意到的人不多。
2021年我们回国的时候，国内这个赛道就是空白，全球范围内也不多。最直接的同行是英国牛津的Sitryx，也是做各种代谢酶，有做同一个通路、但不同上下游的代谢酶，他们也是由免疫代谢领域很出名的学术带头人成立的公司。赛道这么空白，创业初期对外宣讲基本都是在做市场教育、投资者教育。
▎AI科技评论：未来会往肿瘤方向发展吗？
许柯：有可能，但做肿瘤更难。肿瘤的基因突变能力太强，在生存压力下很快能改变自己的代谢——不用糖代谢了，用氨基酸、脂肪，你的用药可能跟不上它变异的速度。做自免就没有这个问题，因为导致疾病的免疫细胞还算是个正常细胞，没有极端的基因突变，无非是它识别的是宿主自己的健康组织，而不是细菌病毒。
▎AI科技评论：MP-5342现在的临床进展如何？
许柯：我们计划2027年上临床一期，主要是健康人的安全性评测。预计2028年进入二期，在病人身上检测安全性和初步有效性。
03
研发方法论：从病人出发，对AI务实
在AI制药追求“预测一切”的浪潮中，默达选择的路径是，从真实病人的多组学数据出发，反向验证，再成药。AI被用在了务实的地方——晶泰负责海量分子设计和合成，默达自建平台分析临床样本，形成紧密互动的研究成果分析验证链。而对于“虚拟病人”和“虚拟细胞”的概念，许柯也保持着罕见的冷静。
▎AI科技评论：默达的研发路径和传统药企有什么不同？
许柯：传统是从细胞、动物往人推。我们反过来，从病人出发。我们分析病人的基因组学、代谢组学、蛋白组学，比如找100个IBD病人共同上调的代谢通路，看是不是代表了这个疾病的驱动因素。然后回到动物身上做基因敲除验证，如果能阻止疾病发展，再针对这个驱动者做抑制剂。我们的药最终是为了治疗患者，而不仅仅是在动物模型中取得良好的实验结果。
▎AI科技评论：AI在默达的研发中扮演什么角色？
许柯：分两块。一块是小分子研发，我们和晶泰开展合作，将AI辅助的分子筛选与设计，以及化合物的合成、测试和优化结合起来。小分子药物研发需要探索非常广阔的化学空间。用于计算筛选的虚拟分子库非常大，有的可达10的9次方量级甚至更大规模，我们不可能把其中的分子一个一个合成出来，再进行实验测试，因此就要借助高效的计算方法，优先找到值得验证的候选分子。
晶泰的AI与计算平台可以结合项目的研发目标对候选分子及其成药相关性质进行预测和综合评估，逐步缩小筛选范围，帮助我们选择更有潜力的分子进入合成与实验验证。这类海量候选分子的评估和优先级排序，很难仅靠人力完成，也是AI在药物发现中能够发挥价值的重要环节。
在我们的合作中，默达主要围绕免疫代谢机制开展LDHA靶点研究与验证，晶泰则依托AI计算平台和高通量机器人实验平台，深度参与化合物筛选、合成、测试和持续优化。筛选出来的候选分子，仍然需要到湿实验室里去验证，拿到数据再feedback回去训练模型，不断更新预测，并指导下一轮分子设计与优化。这种计算与实验相互验证、持续迭代的研发方式，是我认为AI能够在药物发现中发挥实际价值的重要路径。
▎AI科技评论：另一部分是自建的？
许柯：对，另一块是我们自建的临床多组学病人样本分析平台，由鲜于安今博士负责搭建——他之前就在病人数据平台工作。这块针对病人样本做基因组学、代谢组学、蛋白组学等多维度分析，从中找规律，验证靶点，或者找到新的药物靶点。
为什么以前药物研发失败率很高？一个重要原因在于，早期研究往往依赖细胞和动物模型，未能充分考虑人体疾病的复杂性。我们关注的是IBD病人共同的疾病特征：这个特征是不是导致这些人有IBD的原因？往回推，找到共同的通路或靶点，再回到动物身上做验证——基因敲除掉这个通路上的代谢酶，看能否预防疾病；疾病发展时用我们的小分子抑制，验证能否发挥治疗作用。如果能，就说明找到了疾病的驱动者。
从病人出发，比从细胞或动物出发更接近真实的疾病机制和临床需求。
▎AI科技评论：除了AI筛选，其实你们也有考虑类器官，比如和耀速科技合作。
许柯：因为药物研发有个很大的顾虑：从临床前动物模型到人体临床试验，中间有个巨大的鸿沟——动物上看到药效，未必能在人身上看到。怎么减少这个鸿沟？其中一个办法就是用来自病人或健康人的组织样本，做类器官或器官芯片，在进入人体临床试验前，通过简化的人源模型做初步的安全性和药效验证。
我们做自免这块发现，相较于肿瘤领域，市面上自免相关的类器官或器官芯片造模仍处于相对早期的发展阶段。肿瘤领域相关模型已经相对成熟，但免疫相关的更加具有挑战性。因为它需要类器官加上免疫系统，共同组成一个更复杂的共培养体系。我们的免疫背景很强，耀速在器官芯片这块技术很强，如果我们能把多种类器官加上免疫系统共培养，做自免临床前的药效和安全性验证，目前在国内尚不多见。
耀速就在我们隔壁，我和谢博（耀速科技CEO谢鑫）关系挺好的，他们来深圳之前我们就认识了，是科研圈的朋友。刚好他们做这块、也有市场需求，一聊就可以合作。
▎AI科技评论：你如何看待AI在生物医药中的应用前景？
许柯：AI制药未来肯定会有质的飞跃，GPT时刻那种程度。但目前生物不是像数学、物理那样可以定量的。生物太多复杂变数、不确定性。同一个protocol，两个人做出来的结果可能相反。AI需要确定性的input，但生物实验的数据质量很难保证——你甚至不知道它变在哪里、为什么变。所以我认为，现阶段，相较于复杂的生物学预测，我们更看重AI在海量筛选方面的应用价值。
▎AI科技评论：你感受到AI制药的投资水温有什么变化？
许柯：从融资额上就可以看到热度。但这是一个必然经过的过程：有一个hype，就会下来进入冷静期，淘汰那些技术不太行的，然后慢慢有第二波增长曲线——第二波才是生存下来的、有技术的、能做出东西的人。这是正常的过程，一定要大浪淘沙。
雷峰网原创文章，未经授权禁止转载。详情见转载须知。
为了您的账户安全，请验证邮箱
请验证您的邮箱
本文由 AI宝库自动抓取并翻译，版权归原作者所有；如需查看原文请点下方链接。
You are using a low-end IE browser and strongly recommend a faster and safer browser for your Thunder Peak account security and better product experience.
This is a temporary link, only for article preview, which will expire at the time
0
Introduction: “We now value the value of AI's applications in mass selection more than in complex biological predictions.”
The author Chou Lai
Edit Peak
The United States Prix Galien USA, which [editor by] was dubbed "the Nobel Prize for the Medical Community", published the 2026 Award List on 29 October.
It is worth noting that for the second year in a row, the Chongtai technology incubated innovation enterprise has been nominated for the Award. Last year, SIGX1094, the world's premier stomach cancer target, was nominated for "Best Biotechnologies"; this year, Lehman and Moda Biob were nominated for "Best Starters" — the former being nominated for an enhanced metabolic CAR-T therapy at one in a thousand conventional dose, with the help of Jing Tai, and the latter working with Cing Tai to develop a global first-rate oral LDH inhibitorizer MP-5342, targeting a global free market of over $100 billion.
Following the presentation of the award, the Ryboun Science and Technology Review conducted an in-depth dialogue with Higsian, Le Man and Moda Biology, followed by Moda Biology.
In the spring of 2021, Xueko of Cornell Medical School received two invitations. Two well-known American professors in the field of immunization, who each founded an immunization treatment company, were invited to join as co-founders.
Good treatment, clear direction, and everything is ready — Hekke refused.
"Why don't I do my own things?"
At that time, in a joint training project with Cornell and the Sloane-Catelyn Cancer Center, Xueco had spent nearly eight years studying the intersection between immunization and metabolism. He has a set of data on how immune cells change their metabolic base before launching an attack. The potential of this set of data is, in his view, greater than the offer of any ready-made company.
The CTO of the Moda Biology is rare, and at that time, PatientsLikeMe, the largest patient data company in the United States, was the director of data science, with very high pay. But Shikko didn't spend much of his time trying to convince him that it was less than a present, "My data, he looked at it and knew how good it was."
The two men joined in one shot and decided to go home and start a business. Chong Tae-tai was one of their first investors, and when they returned home, they landed directly in his office building and became “building downstairs” innovative partners.
The immune metabolism course, targeted by Shiko, is still in the continuing exploration phase, even abroad. There are many enzymes in the metabolic network, but the development of new medicines for immunocyte metabolism still faces challenges such as selectivity and safety.
Metabolism is too fundamental, and intervention means that the whole body is intermingled and safety risks are unpredictable. However, Shikko has been working in this field since 2013, and he has observed a key phenomenon: partial participation in free-of-disease immunosuppressive cells, which occurs in the process of abnormal activation - increasing glucose use to support cell function.
"Before bad cells do bad things, you have to eat well." That's a metaphor. Modabian hopes to regulate the abnormal immune response by intervening in specific metabolic links and to explore new treatments for its immune and metabolic diseases.
LDHA inhibitor MP-5342, a small oral molecule developed by Murda, corresponds to the bad cell of the “hungry” immune system, reducing the metabolic activity of the immune cell by inhibiting the fermentation of the road, and returning the immune response to a stable state, rather than crushing the immune system to the bottom of the valley, as with traditional immunosuppressants.
This “hunger therapy” brings more than 600 times the safety window, exceeding the industry's conventional safety threshold (usually 40 times). Normal cells do not normally rely on this route, which is needed at a high level only by stress-ridden pathologies. In contrast, the annual sales were close to $20 billion in cardiac risk, and the current JAK inhibitor was a black box warning.
Murda did not start with the biggest silver crumb or special skin disease in the patient ' s base, but instead preferred IBD. Existing drugs for this disease generally face a clinical decomposition rate of less than 40 per cent, and serious patients may need 20 toilets a day, even complications such as intestines. In addition, there is an expanded list of areas where there are still unmet clinical needs, such as primary cholesterol and multiple sclerosis.
Based on the same membrane sub-biology mechanism, Moda organisms have discovered a series of new targets for metabolic protein enzymes, which are at key nodes of the metabolic network and have the potential to expand to a wide range of different adaptive conditions. In addition to self-inflicted diseases, Murda is promoting the first global oral micromeric inhibitor (META-001-PH) for pre-historic hyperfascic hyperplasia (First-in-Class) - a rare and child-prone genetic metabolic disease that accumulates herbic acid in a child, most of which enters kidney failure before adolescence. Shekke told us that there are no special drugs in the country.
On the R & D path, Murda chose a “opposed direction” approach to the traditional approach. The development of traditional medicines is mostly driven from cytology or animal models to humans, and Murda, in turn, from the multi-group data of IBD patients, looks for a common upscaling of metabolic circuits within the group of patients and then back to the animals. The gene knock-down test assesses whether the route is involved in the development of the disease and further develops inhibitors for it.
AI can just help them to move the drug discovery from a “snatch needle” to a “precision navigation”. Jingtai ' s AI+ Robotic Platform, which combines the project ' s research and development objectives, is responsible for the rapid identification of potential candidate drugs that may be combined with specific protein target sites from a mass virtual molecular bank, helping to select more potential molecules for synthesis and experimental validation, and Murda ' s self-built AI Platform for the analysis of genome, metabolic and protein data from clinical sample.
But Hsiuk's attitude to AI is clear: biology is not physics, not zero and one, and there's a huge grey area in the middle; and the same protocol, different people can do different things in different cities. Therefore, in his view, Modeda at this stage values the application of AI in mass screening more than in complex biological predictions.
MP-5342 is scheduled to enter the Healthy person Clinic I in 2027 and the patient IIa Clinical Trial in 2028. Moda Biology is also about to complete a new round of financing to support subsequent clinical development.
In the life cycle of the Ai Pharmaceutical experience, Moda's commitment may have come from a lower-level insistence — to turn nearly a decade of academic accumulation into an accelerated clinical process that benefits patients as soon as possible.
Following are excerpts of the dialogue between the founder of the Moda Biology and CEO-Ray Peaks (WeChat Official Account:Ray Peaks Network) and the AI Science and Technology Review, with unchanging editors:
01
Business start-up options: Return or stay?
In 2021, Xueco completed a joint doctoral training at Cornell School of Medicine and at the Sloane-Catelyn Cancer Memorial Centre, holding key data on the cross-cutting area of immunisation metabolism. When two American professors threw him out of the Initiative Invitation to the Immunization Treatment Enterprise, he chose a more difficult path — the possibility of redefinition of the treatment of self-free diseases by immunisation metabolism.
@AiAiTech: When the US gave you a good deal of treatment and there was a professor who invited you to start a business, why did you choose to go home?
CHHOCO: When we left, we had several special teachers in my field of immunization, and they started their own business, and two professors invited me to join as co-founder to develop immunization-related treatments. It's a great opportunity. I don't have to do it myself. They've already made it up.
But then I felt that what I did was more interesting — instead of choosing to do something that was linked to someone else, why didn't I do something that was mine? I had a very determined idea of starting a business.
How did you convince Dr. Anji?
HS: I've been working on this project before, and I've invited Dr. Anne to join us.
He was already at work, with a very high salary level for PatientsLikeMe, the largest patient data platform in the United States. But it was not a great effort to convince him, and everyone was a researcher, and I did not need to persuade him with affection or anything, so I said, "This is my data, and look at it." I don't have to tell you, you know how good it is and how potential it is. As a scientific researcher, talking in data is the most persuasive.
@AiTech: Why did you go back to Shenzhen?
HSK: There was no particular preference for which city, Shanghai, Suzhou or Shenzhen. The main investor and business partner, Chong Tae-tai, are here, and there are many opportunities for future cooperation that have fallen in Shenzhen. When you land, you start working directly inside Jing Tai.
@AiTech: How did you first meet Jing Tae?
Sheko: The first was an early technology fund called IMO Ventures, and they saw the articles we published and thought we were doing a really good immuno-gravity. It's just that the IMO is also an investor in Jing Tai, who thinks that since we're working on small molecular drugs, and Cing Tai should be able to complement the one in AI, we'll introduce us and Suk-woo, and we'll get together before we go home.
▎AI Science and Technology Review: Is it difficult to adapt from scientists to entrepreneurs?
Sheko: Good for me. The shift from scientists to entrepreneurs requires, beyond scientific research, a growing familiarity with team management, financing and product development, and indeed a number of new topics. I was also in the process of exploring career directions and was willing to try different ways of working, so it was easier to get into a learning-by-doing, adjusting state.
AIT: Have there been any conceptual changes since the CEO?
CHECK: All I've been thinking about is how to make cake big and want to create value through long-term research and development and teamwork, and to value the long-term consensus with investors. This is one of the strongest things I've ever done in my life.
▎AiTechnology Review: Are you targeting Murda, as a platform or as a pipe?
CHUCK: My own understanding is that we are a pipeline company — so-called platforms that serve us to develop pipelines. Platform properties do, but my location is a pipe, mainly for development products.
▎AI Science and Technology Review: What is Moda's most desirable goal for the next 12 months?
HS: It is to advance MP-5342 into clinical and to complete a new round of financing to support subsequent clinical development — drug research and development requires sustained investment.
02
Technical Path: Making bad cells “under-fed”
Moda ' s core technology is based on the immune metabolism studies that Hekke has accumulated since 2013. Unlike the traditional approach to immunosuppression or cell removal, Moda has proposed a “metabolization point” mechanism: to allow immunosuppressive cells that attack their own tissue “to eat less” before “black” and thus restore immunosuppression without causing immuno-deficiency.
@AiAiTech: Can we explain the technical principles of Moda?
HSK: There are many causes of autoimmune diseases, but immunization disorders are at the core of the common: IBD is primarily driven by abnormal T-cell (and inherent immunosuppression) reactions to intestinal strains, while erythalamus is characterized by over-mobilization of B-cells with T-cell-aided and the creation of pathogenic self-antibodies, all of which are excessive immunofunctional disorders and fail to recognize me.
These immunosuppressive cells share a very large common characteristic: before they become bad cells that lead to disease, there will be a process of metabolic reprogramming — it will become bad people, first of all, to change its own metabolic base. Can you imagine how a bad guy can do a bad thing without having the strength to eat? Before the war, weed first. This metabolism is particularly evident in glucose metabolism, where bad cells absorb glucose for energy.
Our strategy is, since you have to eat first to do bad things, I won't let you eat. - By metabolic enzyme inhibitors, crushing its metabolic energy-traffic activity, similar to a “hunger therapy”.
@AiTech: Will this “food loss” starve normal immunosuppressors?
HS: No. The immune system has been dynamically balanced. Immunisation overexposure, but being suppressed too much is similar to immuno-deficiency and is susceptible to bacterial, viral and fungi infections, with a high risk of infection. It's got to be balanced. It can't be excessive or low.
Our treatments are different from those available. These, such as the Adawood Monovalent, JAK Repressants, are a powerful way to suppress all immune capacity, so there is a black box warning, a high risk of infection and a cardiovascular risk. We are putting immunization back into a state of stability and balance, without having to press low.
How will you respond to this security concern about access?
CHECK: This is also the most frequently asked by investors, "Do you have a high risk of interfering with glucose metabolism?" In fact, we've done a lot of experiments to prove that this route is a road under stress, and that cells don't use it very often. The cell is normally aerobic breathing, and glucose enters the line particles to generate a large amount of energy; while the glucose fermentation route is needed only in an aerobic state, when it is intensely moving, and less in a normal state of stability. So inhibition has a relatively limited impact on the functioning and survival of normal cells, but it has shown good inhibition for immunosuppressive cells in pathological states.
▎AI-Technology Review: There's another kind of path that's CAR-T. What's the difference?
Xhuko: The CAR-T therapy, which is used in part for self-resorting studies, attempts to replace the immune system by removing groups of cells such as B cells that express specific antigens. Such treatment and its pre-treatment may entail risks such as infection and require close management. We're going through metabolism, not clearance. Immunocellular cells that are free of disease are normal in themselves, just to identify the wrong target. Let's get it back to balance.
IBD? Silver crumbs or skin inflammation should be more frequent.
HS: Medicine must be related to clinical needs - not only to the number of patients, but also to the effectiveness of existing treatments and unmet clinical needs. First-line treatments for silver crumbs and ad hoc dermatitis have become more mature and effective, with relatively low clinical demand. We therefore prefer to concentrate our limited research and development resources on diseases where clinical needs are more acute.
The total mitigation of the IBD (inflammatory enteropathy) drugs available is less than 40 per cent, and clinical needs are high. It consists of two main types: ulcer ulcer enteritis and Crone disease, which may require 20 toilet visits a day for patients who are slightly more serious, looking for toilets at any time, afraid to go out, diarrhea at all times, severe intestine piercing, and haemorrhage in the digestive tract. The cause is self-immunization -- your immune system attacks your intestinal tract, causing chronic inflammation.
Furthermore, IBD patients need to re-adapt the treatment programme, as the total clinical relief rate is still limited because of the availability of publicly approved drugs, some of which are for a particular patient subgroup and are difficult to meet the treatment needs of all patients. So there is a great need for a drug that is both safe and effective.
▎AiTechnology Review: How did other adaptations go off?
HQ: We have had ad hoc dermatitis, multiple sclerosis, original cholesterol, silver crumbs, rheumatism arthritis on animal models, all of which have worked very well. Multiple sclerosis is an attack by the immune system on neurons, which leads to gradual paralysis of the whole body, a mechanism that is very close to us and different from the organ that is free of attack. There are now more than 100 free diseases, and there are several in each organ. The greatest demand for beds is in IBD; there is also primary cholesterol, multiple sclerosis and still unmet need for treatment.
▎AiTechnology Review: The same target also makes rare tube lines?
HSK: Yes, there is also a pipe line with the same target, different molecules, with primary hypergrassic urine, and a very lethal rare disease among children. These children have congenital genetic defects, a large amount of herbic acid is produced in the liver, and the combination of herbic acid and calcium ion becomes calcium herbate, a kidney stone, which accumulates from the very moment of birth for several months. There are currently no special effects drugs in the country, treatment options are still limited, and some patients are dependent on measures such as hydrosis, drug interventions and rock management, and there are large unmet clinical needs.
For seriously ill patients, kidney failure may develop in childhood or adolescence and require organ transplantation. Since some types of disease are caused by metabolic abnormalities in the liver, some patients are required to undergo liver and kidney transplants, which simply cannot solve the problem. This disease doesn't look immune, but this metabolic enzyme of ours can inhibit the production of heraic acid. It is a multifunctional metabolic enzyme, not only a node of catalytic sugar metabolism, but also a node of catalytic acid. So the same target can be adapted in two directions.
▎AiTech Review: Why are there so few people in the direction of metabolic enzymes?
Shekke: There are many targets to be explored in the metabolic network, especially in the development of medicines for immunocellular cells, and many potentials remain unrecognized. Not technically unacceptable, but rather conservative about the metabolic system, which is perceived to be important and risky to interfere with safety, and the cross-cutting subject of immunisation metabolism is at an early stage, with little attention being paid.
When we returned home in 2021, the national track was blank and not many globally. The most direct colleague is Sitryx in Oxford, United Kingdom, who works as a metabolic enzyme of all kinds, with the same route, but not downstream, and is a company of well-known academic leaders in the field of immuno-gravity. There is such a gap in the course of the race that the initial stage of entrepreneurship is basically market education and investor education.
?AiTechnology Review: Will the future move towards tumors?
CHECK: It's possible, but it's harder to do a tumor. The tumour's genetic mutation is too powerful to change its metabolism very quickly under survival pressure -- without sugar metabolism, with amino acids, fats, your medication may not keep pace with its mutation. There is no problem because the immunocellular cell that causes the disease is still a normal cell, there is no extreme genetic mutation, but it recognizes the host ' s own health tissue, not bacterial viruses.
How is the clinical progress of MP-5342?
Hsu Ko: We plan to go to clinical session in 2027, which will focus on health security assessments. In 2028, phase II is expected to test for safety and initial effectiveness in patients.
03
R & D methodology: From a patient's point of view to AI
In the wave of the AI Pharmaceuticals' pursuit of “forecasting everything”, Modeda chose the path from the real patient's multi-group data to the reverse test and then the drug. AI was used in a practical place — Cing Tai was responsible for mass molecular design and synthesis, and Mota built its own platform to analyse clinical sample, creating a closely interactive chain of analysis and validation of research results. And with regard to the concepts of “virtual patients” and “virtual cells”, Hsuico has maintained a rare calm.
AIT: What's the difference between Moda's R & D path and traditional medicine?
HS: Tradition pushes people from cells, animals. We turn from the sick. We analyze patients' genomics, metabolics, proteomics, for example, finding 100 IBD patients who are co-opt the metabolic circuit, to see if it represents the driver of the disease. Then go back to the animal and do a biopsy, and if it can stop the disease from developing, then do a inhibitor on the driver. Our drugs are ultimately for the treatment of patients, not just for good experimental results in animal models.
@AiTech: What role does AI play in Moda research and development?
Sheko: Two. One is the development of small molecules, and we work with Chongtai to combine AI-assisted molecular screening and design, as well as the synthesis, testing and optimization of compounds. Small molecular drug development requires the exploration of very wide chemical space. The virtual molecular arsenal used to calculate screening is very large, some of which can be up to 10 squares or even larger, and it is impossible for us to synthesize one molecule in one and then experiment with it, so that, with efficient calculations, priority will be given to finding a candidate for validation.
The Jingtai ' s AI and Calculator Platform can predict and assess the relevant properties of candidate molecules and their formulations in a comprehensive manner in conjunction with the project ' s research and development objectives, gradually narrowing the screening range and helping us to select more potential molecules for synthesis and testing. The assessment and prioritization of such large-volume molecules is difficult to do with manpower alone and is an important component of AI ' s value in drug detection.
In our collaboration, Mota has focused on LDHA target research and validation around the Immuno-metabolism Mechanism, and Cinta has been deeply involved in the screening, synthesis, testing and continuous optimization of compounds, based on the AI Calculator Platform and the high-throughput robotic experimental platform. Selected candidate molecules still need to be validated in a wet laboratory, refeedbackedback of the data, updated projections and guided round molecular design and optimization. The way in which these calculations and experiments are validated and developed over time is an important path that I believe AI can play a real value in drug discovery.
The other part was built on its own?
HS: Yeah, the other is our own sample analysis platform for multi-group clinical patients, which was built by Dr. few of you -- he worked on the patient data platform before. This piece is for multi-dimensional analyses of genomics, metabolics, proteins, etc. of patients, from which patterns, targets are verified or new drug targets are found.
Why was there a high rate of drug failure? One important reason is that early research often relies on cell and animal models and fails to take fully into account the complexity of human diseases. We are concerned about the common disease characteristic of IBD patients: is it the reason why they have IBD? Push back, find a common route or target, and then go back to the animal for a test -- the gene knocks out the metabolic enzyme on the road, to see if it can prevent the disease; the disease develops with our small molecular inhibition, to see if it can be treated. If so, it means that the disease driver is found.
From the patient, it is closer to real disease mechanisms and clinical needs than from a cell or animal.
▎AiTechnology Review: In addition to the AI screen, you've also been thinking about sub-organisms, for example, working with speed technology.
CHHOCO: There's a huge gap between pre-clinical animal models and human clinical trials — an animal seeing the effects of the drug, not necessarily in humans. How do we reduce this gap? One approach is to make organ types or organ chips using an tissue sample from a patient or healthy person, and to test initial safety and efficacy through a simplified human source model before entering a human clinical trial.
We have discovered that, compared to the field of tumours, free-standing modelling of related organ types or organ chips is still at a relatively early stage of development. Oncology-related models are relatively mature, but immunization-related are more challenging. Because it requires a kind of organ, together with an immune system, to form a more complex co-culture system. We have a strong immune background and a high-speed field of organ chips, and it is not uncommon in the country at this time if we can produce multiple organs together with the immune system for pre-clinical efficacy and safety.
We have a good relationship with Sheba, and we met before they came to Shenzhen. They do this, they have market needs, and they can work together.
AIT: How do you see the future of AI's application in biomedicine?
HS: The AI Pharmaceuticals are bound to make a qualitative leap in the future, at the GPT level. But at the moment, biology is not as quantifiable as mathematics or physics. Biologicals have too many complex variables, uncertainties. In the same protocol, the result could be the opposite. AI needs certainty input, but the quality of data from biological experiments is difficult to guarantee - - You don't even know where it changes, why it changes. So I think at this stage, we value the application of AI in mass screening more than in complex biological predictions.
▎AiTechnology Review: How do you feel about the water temperature of investments in the AI pharmaceutical industry?
HS: The heat can be seen in the amount of the financing. But it's an inevitable process: one hype comes down to a cooler period, takes out those technologies that aren't good, and then slowly has a second-wave growth curve -- - The second wave is the one that survives, has the skills, can make things. It's a normal process.
Original Ray Peak article, unauthorized reproduction. For further information, see the information below.
For your account security, please verify the mailbox.
Please verify your mailbox.
Auto-fetched and translated by AI Gem Box. All rights belong to the original author.