Insilico AI Drug Shows Lung Function Reversal in Phase 2 Trial

Cracking the Code of Life with AI: Using Frontier AI to Develop Therapies Addressing Diseases and — Ai4

Ai4

Targeting lethal diseases with dual-purpose AI molecules enables biotechs to clear regulatory hurdles while validating systemic age-reversal mechanisms in human trials.

AI-assisted summary of Cracking the Code of Life with AI: Using Frontier AI to Develop Therapies Addressing Diseases and — original source. Check the original for full context.

Source published:

Detailed notes

## Background Published by Ai4 on YouTube on August 14, 2026, this presentation features Dr. Alex Zhavoronkov, founder and CEO of Insilico Medicine. The session covers how generative AI, automated robotic platforms, and domain-specific AI models accelerate target discovery and molecular design to transform drug development and extend human longevity.

## TL;DR Dr. Alex Zhavoronkov critiques traditional pharmaceuticals for failing to extend healthy human lifespan and presents Insilico Medicine's dual-purpose AI drug discovery paradigm. By identifying biological targets that intersect core hallmarks of aging with fatal conditions like Idiopathic Pulmonary Fibrosis (IPF), Insilico discovered and advanced a novel TNIK inhibitor into Phase 2a clinical trials after synthesizing just 79 molecules. Zhavoronkov highlights the strategic advantages of China's automated preclinical ecosystem, discusses the multi-system longevity benefits of GLP-1 agonists, and previews an upcoming AI-driven longevity city development in Pudong.

## Atmosphere & style Zhavoronkov delivers a pragmatic, fast-paced, and refreshingly candid keynote characterized by blunt industry critique and grand technological ambition. He moves effortlessly between dense computational pharmacology data and personal anecdotes—such as using GLP-1s to curb evening cravings—giving the talk an authentic, founder-led energy. The presentation carries an urgent, visionary momentum that makes the convergence of generative AI, automated robotics, and systemic lifespan extension feel immediately tangible.

## Key takeaways - [00:04](https://www.youtube.com/watch?v=1_ibDuLTCfA&t=4s) Traditional pharmaceutical development has failed to produce a single clinically validated therapeutic that extends healthy lifespan in healthy individuals. - [02:00](https://www.youtube.com/watch?v=1_ibDuLTCfA&t=120s) Generative AI provides its greatest efficiency in the preclinical phase (target discovery to candidate nomination), while downstream GLP toxicology and clinical trials remain bound by regulatory timelines. - [02:00](https://www.youtube.com/watch?v=1_ibDuLTCfA&t=120s) AI biotech startups should partner or license assets to major pharmaceutical companies to leverage their established clinical trial infrastructure and regulatory relationships. - [05:18](https://www.youtube.com/watch?v=1_ibDuLTCfA&t=318s) Integrating generative AI with China's dense laboratory infrastructure and talent base accelerates preclinical programs roughly fourfold, cutting up to two years off timelines. - [05:18](https://www.youtube.com/watch?v=1_ibDuLTCfA&t=318s) Targeting dual-purpose biology—focusing on severe, reimbursable diseases linked to fundamental hallmarks of aging—enables a viable 'pipeline in a product' commercial path for longevity therapeutics. - [08:44](https://www.youtube.com/watch?v=1_ibDuLTCfA&t=524s) GLP-1 receptor agonists serve as a real-world blueprint for longevity medicines by simultaneously modulating metabolism, systemic inflammation, and psychological impulses. - [11:35](https://www.youtube.com/watch?v=1_ibDuLTCfA&t=695s) Insilico used generative AI to discover TNIK as a novel target regulating four key signaling cascades (TGF-β, TNF-α, NF-κB, and Wnt) for Idiopathic Pulmonary Fibrosis (IPF). - [15:30](https://www.youtube.com/watch?v=1_ibDuLTCfA&t=930s) Generative chemistry optimized the TNIK program with only 79 synthesized small molecules, drastically cutting the high cost and volume typical of traditional hit-to-lead campaigns. - [15:30](https://www.youtube.com/watch?v=1_ibDuLTCfA&t=930s) The TNIK inhibitor demonstrated potent senomorphic effects superior to metformin and rapamycin, showing preliminary reversal of forced vital capacity decline in a Phase 2a clinical trial. - [18:50](https://www.youtube.com/watch?v=1_ibDuLTCfA&t=1130s) Insilico is co-developing an integrated AI Longevity City in Pudong, Shanghai, combining residential spaces, automated labs, and longitudinal health tracking for proactive care by 2028.

## Detailed walkthrough ### [00:04](https://www.youtube.com/watch?v=1_ibDuLTCfA&t=4s) The Grand Frustration with Pharma and the Longevity Mission

Alex Zhavoronkov, founder and CEO of Insilico Medicine, opens his talk with an invitation for researchers and technologists to enter the aging and longevity field. With modern frontier AI and large language models, the barrier to entry has dropped dramatically: "If you don't know how to do it, you can ask your favorite LLM." He prefaces his technical remarks with dual disclaimers—advising the audience not to trade securities based on his words because Insilico is a public company, and warning them never to take unapproved drugs on his advice since he is not a medical doctor and the therapeutics discussed are strictly investigational.

Zhavoronkov shares his overarching frustration with modern biomedical history: despite roughly 150 years of pharmaceutical discovery, the industry has failed to produce a single clinically validated medicine that can grant an extra year of healthy life to an already optimized, healthy person. Regardless of wealth, intelligence, or proactive health tracking, there is no drug on the market with human clinical trial evidence proving age reversal or general lifespan extension in healthy cohorts.

``` TRADITIONAL PHARMA PARADIGM +-------------------------------------------------------------+ | Treats isolated clinical diseases (reactive medicine) | | Fails to add healthy lifespan to non-diseased individuals | +-------------------------------------------------------------+ v DUAL-PURPOSE AI PARADIGM +-------------------------------------------------------------+ | AI Target Discovery (Aging Clocks + Hallmarks of Aging) | | v v | | Primary Approval: Severe Disease Secondary: Longevity | +-------------------------------------------------------------+ ```

[02:00](https://www.youtube.com/watch?v=1_ibDuLTCfA&t=120s) Navigating the Drug Discovery Pipeline and the Preclinical Sweet Spot

Building a sustainable business model in longevity biotechnology requires functioning as a tier-one pharmaceutical engine that routinely produces high-quality drug candidates. Insilico operates as a global enterprise across Boston, Montreal, Abu Dhabi, and Greater China. Zhavoronkov outlines the standard stages of the pharmaceutical value chain, from identifying the biological "protein criminal" driving a disease to engineering molecular "handcuffs" (small molecules or antibodies) to shut it down.

Generative AI provides its most transformative leverage in the early, unregulated territory spanning target discovery to the nomination of a Preclinical Candidate (PCC) or Developmental Candidate (DC). At Insilico, nominating a candidate requires passing non-GLP toxicity studies in at least two species (such as rat and dog, or mouse, rat, and monkey). Beyond candidate nomination, development enters heavily regulated phases: 28-day GLP toxicology packages take 9 to 12 months, after which clinical trials advance "with the speed of traffic." Because established pharma giants possess superior clinical trial infrastructure, investigator networks, and regulatory relationships, AI biotech startups should ideally partner or out-license their clinical assets rather than attempt late-stage development in isolation.

- **Preclinical Sweet Spot:** AI grants exponential speed and cost reduction in target identification, hit-to-lead, and lead optimization. - **GLP Toxicology Bottleneck:** Regulated non-clinical safety studies take 9–12 months regardless of computational speed. - **Commercialization Strategy:** License advanced pre-clinical and early clinical assets to major pharma partners to leverage their hospital networks and trial execution capabilities.

[05:18](https://www.youtube.com/watch?v=1_ibDuLTCfA&t=318s) The China Acceleration Advantage and "Pipeline in a Product" Strategy

Zhavoronkov highlights China as an unmatched operational base for preclinical drug discovery due to its deep talent pool, integrated laboratory infrastructure, industrious work culture, and supportive regulatory environment. Integrating AI with the Chinese discovery ecosystem accelerates programs roughly fourfold, shaving up to two years off standard preclinical timelines. Insilico chose to operate directly within China to "compete in China with China," matching the raw operational execution of local pharmaceutical firms while differentiating on algorithmic biological novelty.

A foundational pillar of Insilico's strategy is developing dual-purpose therapeutics. Instead of seeking an unviable direct regulatory path for "aging," the team identifies targets that are simultaneously implicated in the core hallmarks of aging and specific lethal diseases. The commercial goal is to secure regulatory approval and reimbursement for an acute or chronic disease, making the anti-aging benefits an integrated byproduct: "You kind of do free aging research by aligning yourself with a traditional pharmaceutical drug discovery and development paradigm." This mirrors a "pipeline in a product" model—analogous to aspirin—where a single molecule targeting a central node can eventually expand across multiple indications.

[08:44](https://www.youtube.com/watch?v=1_ibDuLTCfA&t=524s) GLP-1s as Longevity Trailblazers and the ARD Conference

Longevity biotechnology is already hidden in plain sight within major pharmaceutical categories, including cardiometabolic diseases, central nervous system disorders, fibrosis, and chronic inflammation. Zhavoronkov points to the annual Aging Research and Drug Discovery (ARD) conference—which he founded, transitioning from Basel to Copenhagen, and heading to the David Rubenstein Treehouse at Harvard—as a central nexus for these discussions.

At the previous year's event, Andrew Adams (Head of Genetic Medicine at Eli Lilly) and Lotte Bjerre Knudsen (Novo Nordisk leader instrumental in GLP-1 development) presented real-world evidence demonstrating that GLP-1 receptor agonists like semaglutide function as genuine longevity medicines. By modulating metabolism, systemic inflammation, and even psychological compulsion, GLP-1s reduce the incidence of diverse age-related conditions. Zhavoronkov admits to taking a low-dose GLP-1 himself, remarking that it "kills all the demons in my head that want me to eat in the evening" while reducing stress, establishing a clear blueprint for scaling future longevity therapeutics through metabolic and systemic modulation.

``` GLP-1 RECEPTOR AGONIST MULTI-SYSTEM IMPACT +--------------------------------------------------------+ | Central Appetite & Impulse Control (Psychology/Stress) | +---------------------------+----------------------------+ | +---------------------------v----------------------------+ | Systemic Metabolic Regulation (Insulin/Glucose Homeo.) | +---------------------------+----------------------------+ | +---------------------------v----------------------------+ | Organ-Level Disease Reduction (Cardio, Renal, Fibrosis)| +--------------------------------------------------------+ ```

[11:35](https://www.youtube.com/watch?v=1_ibDuLTCfA&t=695s) AI Target Mapping and the TNIK Discovery Campaign

In a 2022 landmark publication, Insilico demonstrated how generative AI can map therapeutic targets against the biological hallmarks of aging, ranking candidates by novelty, druggability, commercial tractability, and pathway intersection. In 2019, backed by a $37 million venture round, Insilico embarked on what statistically seemed like a "mission impossible" with a sub-1% chance of success: discovering a novel biological target and designing a novel small molecule for a fatal age-associated indication from scratch, and advancing it through human clinical trials.

The AI platform pinpointed TNIK (TRAF2- and NCK-interacting kinase), an intracellular target that simultaneously regulates four major biological signaling cascades: TGF-β, TNF-α, NF-κB, and Wnt. Insilico selected Idiopathic Pulmonary Fibrosis (IPF) as the lead indication. IPF is an aggressive, age-driven disease with an average onset around 65 years of age. While healthy adults naturally lose 30 to 40 mL of forced vital capacity (FVC) annually after 65, untreated IPF patients suffer catastrophic pulmonary decline of roughly 400 mL per year, with very poor existing standard-of-care options.

- **Target Discovered:** TNIK (TRAF2 and NCK-interacting kinase). - **Pathways Modulated:** Downregulates TGF-β, TNF-α, NF-κB, and Wnt signaling networks. - **Primary Indication:** Idiopathic Pulmonary Fibrosis (IPF). - **Clinical Biomarker:** Forced Vital Capacity (FVC) measurement.

[15:30](https://www.youtube.com/watch?v=1_ibDuLTCfA&t=930s) Robotic Screening, Senomorphic Efficacy, and Phase 2 Clinical Trials

In March 2024, Insilico published the end-to-end journey of the TNIK program in *Nature Biotechnology*. Demonstrating unprecedented medicinal chemistry efficiency, the team synthesized only 79 small molecules during the entire campaign—a fraction of the thousands typically generated by legacy pharma programs, where testing each candidate molecule costs $2,500 to $5,000. Testing in fully automated robotics laboratories built by Insilico in Suzhou and Shanghai demonstrated that the lead molecule possessed potent senomorphic properties, clearing and reversing cellular senescence in both drug-induced and replicative models more effectively than standard longevity benchmarks metformin and rapamycin.

``` TNIK PROGRAM TIMELINE +-------------------------------------------------------------+ | AI Target Discovery: TNIK identified as anti-fibrotic target| +------------------------------+------------------------------+ | +------------------------------v------------------------------+ | AI Chemistry: 79 molecules synthesized (Hit-to-Candidate) | +------------------------------+------------------------------+ | +------------------------------v------------------------------+ | Fully Automated Robotics Validation (Suzhou & Shanghai Labs)| +------------------------------+------------------------------+ | +------------------------------v------------------------------+ | Phase 1 Safety Trials (Healthy Volunteers, US & China) | +------------------------------+------------------------------+ | +------------------------------v------------------------------+ | Phase 2a Double-Blind Trial (n=71, 12 weeks, Nature Medicine)| | * Reversal of FVC decline observed at high dose | | * Proteomic blood profiling + 6 biological aging clocks | +-------------------------------------------------------------+ ```

The clinical program advanced into parallel Phase 2a trials in the United States and China. The China double-blind, placebo-controlled trial enrolled 71 patients over 12 weeks, with results published in *Nature Medicine*. At high doses, the drug demonstrated not just a deceleration of lung function decline, but an early, statistically promising reversal of forced vital capacity decline. Insilico gathered comprehensive longitudinal proteomics data from patients throughout the trial and evaluated six distinct biological aging clocks to quantify systemic age reversal in human clinical data.

[18:50](https://www.youtube.com/watch?v=1_ibDuLTCfA&t=1130s) The 2028 Pudong AI Longevity City Project

Looking toward the convergence of foundation AI models and human biology, Zhavoronkov recalls a 2021 collaboration with AI pioneer Dr. Kai-Fu Lee and longevity physician Dr. Evelyne Bischof on training multimodal generative "life models" that understand a patient's biological trajectory from birth to death.

Applying these models to the built environment, Insilico is co-developing an integrated urban longevity community in Pudong, Shanghai, slated to open in Q3 2028. Designed as an architectural and scientific hub, the complex combines residential living, commercial districts, retail, a five-star hotel, advanced robotics, and dense clusters of biotechnology labs. Residents can opt in to share real-time health data with on-site biomedical researchers and an integrated clinical facility designed to provide proactive diagnostics and ensure "the hospital will not let you die of something stupid." Zhavoronkov envisions the visible presence of automated labs and humanoids inspiring the next generation away from social media distractions and toward careers in longevity and biotechnology.

## Notable quotes > [00:04](https://www.youtube.com/watch?v=1_ibDuLTCfA&t=4s) If you don't know how to do it, you can ask your favorite LLM. — Alex Zhavoronkov

> [05:18](https://www.youtube.com/watch?v=1_ibDuLTCfA&t=318s) You kind of do free aging research by aligning yourself with a traditional pharmaceutical drug discovery and development paradigm. — Alex Zhavoronkov

> [08:44](https://www.youtube.com/watch?v=1_ibDuLTCfA&t=524s) kills all the demons in my head that want me to eat in the evening — Alex Zhavoronkov

> [18:50](https://www.youtube.com/watch?v=1_ibDuLTCfA&t=1130s) the hospital will not let you die of something stupid. — Alex Zhavoronkov

## Mentioned Alex Zhavoronkov, Insilico Medicine, Science MMAI Gym, Aging Research and Drug Discovery (ARD) conference, Andrew Adams, Eli Lilly, Lotte Bjerre Knudsen, Novo Nordisk, GLP-1, TNIK, Idiopathic Pulmonary Fibrosis (IPF), Nature Biotechnology, Nature Medicine, Kai-Fu Lee, Evelyne Bischof, Pudong, Shanghai

## Video description Alex Zhavoronkov - Founder and CEO, Insilico Medicine

How will generative AI and automation transform the pharmaceutical industry? In this presentation, Dr. Alex Zhavoronkov, CEO of Insilico Medicine, will detail how generative AI, automated platforms, and AI agents will accelerate the discovery of novel biological targets and the design of new molecular structures. He will explain how reinforcement learning and end-to-end automation will enable drug candidates to move into human clinical trials in record time while significantly reducing the cost and duration of the preclinical phase. The talk will also show how these AI capabilities will extend beyond traditional drug discovery into aging research, agricultural chemistry, and environmental sustainability. He will also introduce Science MMAI Gym, a domain-specific training environment designed to transform any causal or frontier Large Language Model (LLM) into a high-performance engine for real-world drug discovery and development tasks.Overall, the session will present a benchmark for the current and emerging capabilities of end-to-end AI in biotechnology and its potential to extend human longevity.

[Source](https://www.youtube.com/watch?v=1_ibDuLTCfA)

中文简报

Insilico AI药物在2期临床中展现肺功能逆转 利用双效AI分子靶向致命疾病，使生物科技公司得以跨越监管障碍，同时在人体试验中验证全身性衰老逆转机制。 Alex Zhavoronkov 博士批评传统制药未能延长人类健康寿命，并介绍了 Insilico Medicine 的双效 AI 药物研发范式。通过寻找连接衰老核心特征与特发性肺纤维化（IPF）等致命疾病的生物靶点，Insilico 仅合成了79个分子便发现并推动一款新型 TNIK 抑制剂进入 2a 期临床试验。Zhavoronkov 重点指出了中国自动化临床前生态系统的战略优势，探讨了 GLP-1 受体激动剂在多系统上的长寿益处，并预告了即将在浦东开展的 AI 驱动长寿城市开发项目。 生成式 AI 在临床前阶段（靶点发现至候选分子确立）展现出指数级的提速降本能力，但后续 GLP 毒理实验与临床阶段仍受制于刚性的监管时间窗口。 (2:00) 将生成式 AI 与中国密集高效的实验室基建和人才网络结合，可使临床前开发速度提升约 4 倍，缩短多达 2 年的研发周期。 (5:18) 长寿药物商业化应借鉴 GLP-1 模式，采取“单产品多管线”策略，主攻兼具衰老底层特征与高致死率、高报销价值的重症适应症。 (8:44) 英矽智能（Insilico Medicine）利用 AI 发现能调控四大信号通路的 IPF 全新靶点 TNIK，仅合成 79 个分子即完成先导化合物优化，并在 2 期临床中观察到患者肺功能下降趋势的逆转迹象。 (15:30)

Reading summary

Alex Zhavoronkov, founder and CEO of Insilico Medicine, argues that generative AI enables a dual-purpose drug discovery model that treats severe, age-associated diseases while simultaneously targeting underlying biological aging mechanisms. The Dual-Purpose Route to Longevity Therapeutics (0:04) Zhavoronkov notes that roughly 150 years of pharmaceutical development have failed to deliver a single clinically validated drug that adds an extra healthy year of life to an already healthy individual. Because regulatory agencies lack an approved indication for aging itself, biotech companies cannot run direct commercial trials for general lifespan extension. Insilico's strategy addresses this by discovering targets that operate at the intersection of fundamental aging hallmarks and fatal clinical conditions, effectively funding aging research through the traditional pharmaceutical reimbursement system. This approach mirrors the emerging role of GLP-1 receptor agonists. As highlighted by researchers from Eli Lilly and Novo Nordisk at the Aging Research and Drug Discovery (ARD) conference, GLP-1 therapies demonstrate how addressing systemic inflammation, metabolic signaling, and even behavioral compulsions can reduce multiple age-related chronic disorders simultaneously. Insilico aims to treat its pipeline candidates as broad systemic platforms—analogous to aspirin—securing initial approval for severe diseases before expanding into wider indications. Preclinical AI Acceleration and Commercial Strategy (2:00) The primary operational leverage for generative AI occurs in the unregulated early stages of discovery: target identification, hit-to-lead development, and preclinical candidate nomination. Candidate nomination at Insilico requires non-GLP safety validation across at least two animal species. Beyond this point, development collides with regulatory bottlenecks, including mandatory 28-day GLP toxicology programs that take 9 to 12 months regardless of computational speed. To maximize efficiency during early development, Insilico established core operations in China. Zhavoronkov claims that pairing AI algorithms with China's laboratory infrastructure, technical workforce, and regulatory environment accelerates preclinical development roughly fourfold, eliminating up to two years compared to Western timelines. Because late-stage clinical trials advance slowly and require vast clinical networks, Insilico's business model relies on licensing advanced assets to major pharmaceutical corporations rather than funding Phase 3 trials independently. TNIK Validation from Discovery to Phase 2 (11:35) To validate its generative AI platform, Insilico initiated an end-to-end program targeting Idiopathic Pulmonary Fibrosis (IPF), an aggressive age-linked disease where lung capacity declines by roughly 400 mL annually compared to 30 to 40 mL in normal aging. The platform identified TNIK (TRAF2- and NCK-interacting kinase), an intracellular node that regulates four major pathways: TGF-β, TNF-α, NF-κB, and Wnt. The team synthesized only 79 small molecules to reach a candidate—a fraction of the thousands typical in traditional chemistry, where screening costs $2,500 to $5,000 per compound. Testing in automated robotic facilities in Suzhou and Shanghai demonstrated that the compound cleared and reversed cellular senescence more effectively than benchmark longevity molecules metformin and rapamycin. The program progressed to a 12-week, double-blind Phase 2a trial in China with 71 patients, published in Nature Medicine, which demonstrated early, dose-dependent reversal of forced vital capacity decline alongside systemic profiling across six biological aging clocks. Multimodal Life Models and Urban Longevity Projects (18:50) Zhavoronkov envisions broadening biological AI from narrow molecular design tools into multimodal 'life models'—a concept developed with Dr. Kai-Fu Lee and Dr. Evelyne Bischof—that model health trajectories from birth to death. This philosophy extends beyond pharmacology into the built environment through a planned urban longevity development in Pudong, Shanghai, scheduled to open in the third quarter of 2028. The project integrates residential housing, commercial spaces, and a five-star hotel with automated robotics centers and biotechnology laboratories. Under this model, residents can opt to supply longitudinal health data to on-site clinical facilities for real-time diagnostic screening and preventative treatment, while visible laboratory operations aim to cultivate public and youth engagement in biotechnology. While Insilico's computational pipeline and China-based laboratory automation demonstrate marked preclinical efficiencies, proving that dual-purpose molecules can slow human biological aging hinges on ongoing Phase 2 clinical outcomes and subsequent late-stage partnership execution.

精炼阅读摘要

英矽智能（Insilico Medicine）创始人兼首席执行官Alex Zhavoronkov指出，传统制药业长期局限于单一疾病的被动治疗，未能提供任何经临床验证可延长健康人群寿命的药物；为此，英矽智能利用前沿生成式AI开创“双重用途”研发范式，将针对致死性疾病的管线研发与抗衰老机制相结合，以重塑生物医药的研发效率与商业落地路径。 传统制药困境与双重用途研发策略 (0:04) Zhavoronkov指出，过去约150年的制药史中，行业始终专注于疾病发生后的被动治疗，未能研发出任何一款在临床试验中被证实能使健康人群健康寿命延长一年的药物。受制于现行监管体系，“衰老”本身无法直接作为药物审批的临床适应症，因此长寿科技企业必须采用“双重用途”（Dual-Purpose）策略。该模式主张利用AI算法寻找既处于衰老核心标志物网络、又直接参与严重致死性疾病的靶点，通过治疗特定急慢性疾病获得监管批准与医保支付，将抗衰老益处转化为合规研发的副产品。 Zhavoronkov强调，这一逻辑类似于阿司匹林等多适应症药物的“单产品管线”（pipeline in a product）模式。类似机制已在心血管与代谢领域初现端倪，例如GLP-1受体激动剂（如司美格鲁肽）不仅通过中枢神经调节食欲与冲动，还能改善系统性代谢稳态、减轻慢性炎症及纤维化，在真实世界数据中展现出跨多器官系统的长寿药物特征。 临床前研发杠杆与跨国协同布局 (2:00) 在药物研发价值链中，生成式AI的最大效能杠杆集中在早期非监管阶段，即从靶点发现到临床前候选药物（PCC）或开发候选药物（DC）的提名。按照英矽智能的标准，候选药物提名必须在至少两种动物（如大鼠与犬类，或小鼠、大鼠与猴）中通过非GLP毒理学评估。然而，一旦进入受监管环节，算力优势将面临现实周期的硬性约束：28天的GLP规范毒理研究通常不可避免地耗时9至12个月，临床试验推进速度也完全依赖外部执行环境。因此，Zhavoronkov主张AI初创企业应将后期临床资产对外授权或与跨国大药企合作，以借助其庞大的医院网络与临床开发能力。 为最大化早期研发效率，英矽智能将全球业务布局于波士顿、蒙特利尔、阿布扎比及大中华区，并以中国作为临床前研发的核心执行基地。Zhavoronkov表示，中国拥有深厚的人才储备、一体化的实验室基础设施与高效的工作文化，将AI算法与当地研发体系结合可使临床前开发速度提升约4倍，缩短多达两年的研发周期，使企业在算法生物学创新与工程执行力两端建立竞争壁垒。 TNIK靶点发现与肺纤维化临床验证 (11:35) 2019年，英矽智能在一轮3700万美元的风险投资支持下启动了一项被视作成功率低于1%的实验项目：完全由AI从头发现新靶点并设计全新小分子化合物。AI平台最终锁定了TNIK（TRAF2与NCK相互作用激酶），该靶点能同时下调TGF-β、TNF-α、NF-κB和Wnt四大核心信号通路。针对该靶点，团队选定特发性肺纤维化（IPF）作为首发适应症。健康成年人在65岁后用力肺活量（FVC）每年生理性下降30至40毫升，而IPF患者每年急剧丧失约400毫升，临床亟需突破性治疗手段。 该项目的全流程研发成果于2024年3月发表于《Nature Biotechnology》。得益于生成式化学算法，该管线在整个筛选周期中仅合成了79个小分子候选物，远低于传统药企合成数千个分子的规模（传统模式下单分子测试成本约为2500至5000美元）。在苏州与上海的自动化机器人实验室验证中，候选分子在复制型与药物诱导型细胞模型中展现出优于二甲双胍与雷帕霉素的衰老表型改善能力（senomorphic properties）。随后在发表于《Nature Medicine》的中美双盲安慰剂对照2a期临床试验中（中国队列包含71名患者，为期12周），高剂量组不仅减缓了肺功能恶化，还显示出力度积极的FVC逆转信号，并通过纵向蛋白质组学及6个生物衰老时钟对患者机体系统性年轻化进行了临床量化。 多模态生命模型与浦东长寿社区项目 (18:50) 面向生命科学的未来融合，Zhavoronkov回顾了其在2021年与李开复博士及长寿医学专家Evelyne Bischof博士的合作构想，即利用多模态生成式AI训练覆盖个体从出生到死亡全生物轨迹的“生命模型”（life models）。以此为理论延伸，英矽智能正参与在上海浦东共建一座预计于2028年第三季度启用的城市AI长寿综合社区。 该园区规划融合了居住区、商业设施、五星级酒店以及高密度的自动化机器人与生物技术实验室。入驻居民可在知情同意下选择与现场医疗科研团队共享实时健康数据，依托内置临床设施进行前瞻性预防与主动诊断，旨在防范非必要的早逝风险。Zhavoronkov预期，将自动化研发实体与人形机器人融入日常物理空间，有助于引导下一代减少对社交媒体的沉迷，转向长寿科技与生物医药领域的深度探索。 Zhavoronkov描绘了以AI驱动跨系统靶点发现与长寿医学落地的蓝图，但其核心候选药物仍处于临床试验阶段，且算法无法绕过长达数月乃至数年的GLP毒理与后期临床试验周期；该范式能否从特定纤维化适应症成功推广为普适性的延寿疗法，仍需更大规模的多中心临床数据与长周期验证。

Detailed notes — zh

## Background 本内容由 YouTube 频道 Ai4 于 2026 年 8 月 14 日发布。视频记录了英矽智能（Insilico Medicine）创始人兼首席执行官 Alex Zhavoronkov 博士的主题演讲，深入探讨如何利用生成式 AI、自动化平台及 AI Agent 加速全新靶点发现和分子设计，并将前沿 AI 能力拓展至抗衰老研究与生命延长领域。 ## TL;DR Alex Zhavoronkov 指出传统制药业百年来缺乏直接验证的延寿药物，阐明了利用生成式 AI 探索“双重用途”（同时治疗疾病与延缓衰老）靶点的研发范式。他分享了英矽智能如何结合中国高效研发基建与自动化机器人实验室，仅合成 79 个分子便成功研发出针对特发性肺纤维化（IPF）的全新 TNIK 抑制剂，并在 2 期临床试验中展现出逆转肺活量衰退与生物学年龄的积极信号。最后，他展望了涵盖多模态数据的“生命模型”，并公布了计划于 2028 年在上海浦东交付的长寿智慧社区愿景。 ## Atmosphere & style 演讲节奏明快且充满极客雄心与科学严谨感。Zhavoronkov 以务实而坦诚的创业者姿态开场，穿插个人服用 GLP-1 药物的生动体验，使硬核的生物医药数据与前沿算法讲解显得引人入胜。整场分享从对传统制药痛点的深刻反思，层层递进至颠覆性临床成果展示与未来智慧长寿社区的宏伟蓝图，带给观众一种见证生命科技与人工智能交汇落地的澎湃未来感。 ## Key takeaways - [00:04](https://www.youtube.com/watch?v=1_ibDuLTCfA&t=4s) 全球制药工业 150 年来尚未在临床上证实有药物可为健康人延长寿命或逆转生物学年龄，而大模型等 AI 工具正显著降低抗衰老研究的准入门槛。 - [02:00](https://www.youtube.com/watch?v=1_ibDuLTCfA&t=120s) 从项目立项到确定临床前候选化合物（PCC）是 AI 赋能的最大自由探索区间，AI 药企通过对外授权（License-out）将后期临床交由传统大药企是最优商业闭环。 - [05:18](https://www.youtube.com/watch?v=1_ibDuLTCfA&t=318s) 中国高度集成的供应链基础设施可将临床前阶段缩短约 2 年，结合端到端 AI 赋能可使整体药物研发提速近 4 倍。 - [05:18](https://www.youtube.com/watch?v=1_ibDuLTCfA&t=318s) 推行“双重用途”（Dual-purpose）靶点策略，挑选同时调控疾病与衰老机制的靶点，先通过特定疾病的传统审批通道获批上市，再拓展至延缓衰老领域。 - [08:44](https://www.youtube.com/watch?v=1_ibDuLTCfA&t=524s) 跨国药企出示的真实世界数据表明 GLP-1 药物兼具全身代谢调控与衰老相关疾病预防效果，为双重用途抗衰药物确立了可行的商业与临床范式。 - [11:35](https://www.youtube.com/watch?v=1_ibDuLTCfA&t=695s) 纯 AI 驱动从零发现了可同时调控四条核心通路的全新靶点 TNIK，并选定高死亡率的衰老驱动型疾病特发性肺纤维化（IPF）作为首发适应症。 - [15:09](https://www.youtube.com/watch?v=1_ibDuLTCfA&t=909s) 依托生成式 AI 的极高命中率，该项目仅合成 79 个分子即锁定候选药，并在全自动化生物机器人实验室中证实其抗衰老性能优于二甲双胍和雷帕霉素。 - [15:09](https://www.youtube.com/watch?v=1_ibDuLTCfA&t=909s) TNIK 抑制剂在 2 期临床试验中使患者肺功能呈现出改善与逆转的早期信号，并在 6 种衰老生物钟评估中初步呈现出生物学年龄逆转的迹象。 - [18:50](https://www.youtube.com/watch?v=1_ibDuLTCfA&t=1130s) 基于训练人类全周期多模态生物数据的“生命模型”愿景，英矽智能正于上海浦东建设集科研、居住与主动健康防御于一体的 2028 年长寿智慧社区。 ## Detailed walkthrough ### [00:04](https://www.youtube.com/watch?v=1_ibDuLTCfA&t=4s) 投身抗衰老研究的契机与制药业百年的“长寿困局” 英矽智能（Insilico Medicine）创始人兼首席执行官 Alex Zhavoronkov 在演讲开场时呼吁，当下应当有更多研究人员利用前沿人工智能投身衰老与长寿研究。得益于大语言模型（LLM）等 AI 工具的普及，该领域的准入门槛已大幅降低。Zhavoronkov 随后给出了合规免责声明：公司已上市，演讲内容不构成证券买卖建议；同时他强调自己不是执业医生，所讨论的内容均属于前沿实验性探索，切勿擅自以此作为用药依据。 谈及投身该领域的根本动力，Zhavoronkov 指出全球制药行业过去 150 年来面临的一个严峻现实：整个医药工业至今未能研发出哪怕一种能够在临床上明确证实“可为健康人额外延长一年寿命”或实现生物学年龄逆转的药物。哪怕一个人极其富有、认知前沿且定期接受高端体检与健康优化，只要没有患上特定疾病，现阶段服用的任何抗衰药物都缺乏严格的人体临床试验有效性证据。发现针对衰老本身的靶向药物极其艰难，但这也正是必须攻克的方向。 ```mermaid flowchart TD accTitle: 双重用途药物研发与长寿验证路径 subgraph S1["传统监管审批通道 (疾病治疗)"] A["AI 靶点发现与筛选<br/>(衰老标志物 / 生物钟)"] --> B["设计双重用途分子<br/>(例如 TNIK 抑制剂)"] B --> C["针对特定疾病临床试验<br/>(例如 IPF 特发性肺纤维化)"] C --> D["获批上市与适应症拓展"] end subgraph S2["长寿转化通道 (抗衰老获益)"] D --> E["真实世界数据与长寿验证"] C -.->|"收集蛋白质组学数据"| F["衰老生物钟检测与评估"] F --> E end ``` --- ### [02:00](https://www.youtube.com/watch?v=1_ibDuLTCfA&t=120s) Insilico 的商业闭环与 AI 在临床前研发中的“超能力” 为了支撑漫长且耗资巨大的长寿药物研发，Zhavoronkov 经历了几度濒临资金耗尽、再到重新充裕的创业起伏。他强调，单纯追求长寿概念无法维持长期运营，必须建立自我造血的可持续商业模式——即在全球制药工业中提供顶尖的药物发现技术服务与高质量候选药物管线。目前英矽智能已建立起覆盖波士顿、蒙特利尔、阿布扎比和大中华区的全球化研发布局。 在整个制药价值链中，从项目立项到确定临床前候选化合物（PCC，即开发候选物 Developmental Candidate）这一阶段，是 AI 能够赋予药企最大“超能力”的无监管自由探索区间。传统的药物研发从识别致病的“蛋白质罪犯”（靶点发现）到为其量身定制小分子或抗体“手铐”（分子设计），往往要耗费数年时间与数千万美元。 - **候选物提名标准**：英矽智能内部对开发候选物的要求是必须完成至少两个物种（如大鼠与犬类，或小鼠/大鼠与猴类）的非 GLP 毒理学测试，以证实其基本安全性。 - **监管与临床分工**：一旦提名候选分子，进入 28 天 GLP 毒理实验阶段（通常耗时 9 至 12 个月）及后续的人体临床试验后，研发就必须按照监管部门的既定节奏推进。 - **商业化分工策略**：传统跨国大型药企在医院渠道、主要研究者（PI）资源和监管沟通上具备深厚积累；因此在理想模式下，AI 制药公司在完成早期发现与候选物验证后，应优先将管线授权（License-out）给传统制药巨头推进后期临床，而非独自承担所有临床负担。 --- ### [05:18](https://www.youtube.com/watch?v=1_ibDuLTCfA&t=318s) 中国研发平台的加速度与“双重用途”靶点选择 Zhavoronkov 分析了当今生物医药领域的新格局：中国凭借高度集成的供应链基础设施、庞大且勤勉的高端研发人才储备以及支持性政策，已成为极其高效的临床前药物研发平台。单纯依托中国的这套高效研发体系，就能在临床前阶段省下约两年的时间；若再叠加端到端 AI 技术的赋能，整个药物发现与临床前开发进程可提速近 4 倍。 面对中国本土传统药企在研发执行速度上的强大优势，英矽智能选择直接深入中国市场展开竞争。本土药企的专长在于极致的开发速度，而英矽智能则在分子与靶点的新颖度（Novelty）上构筑差异化壁垒。 在靶点选择逻辑上，英矽智能采取了独特的“双重用途”（Dual-purpose）策略： - **疾病与衰老同源**：挑选出的治疗靶点必须同时参与人类疾病进展与底层衰老机制，且需要同时获得人类与动物的衰老生物钟（Aging Clocks）数据及衰老标志物（Hallmarks of Aging）的交叉验证。 - **以病养衰的获批路径**：首先将药物按照传统制药监管范式针对某一严重疾病推进临床试验并争取获批；一旦药物正式上市并被证实兼具延缓衰老的效果，就相当于“免费完成了抗衰老药物的开发”。 - **“产品即管线”（Pipeline in a Product）理念**：类似于阿司匹林能够同时用于止痛、退烧和抗凝血，研发团队致力于寻找能针对多个适应症起效的核心全新靶点，先攻克原发适应症，再将适应症逐步外延至其他老年病及衰老过程。 --- ### [08:44](https://www.youtube.com/watch?v=1_ibDuLTCfA&t=524s) 传统药企的长寿切入点与 GLP-1 药物带来的范式突破 长寿生物技术在传统大型制药工业中并非凭空存在，而是具体映射在各大疾病领域中，包括心血管代谢（Cardiometabolism）、中枢神经系统退行性疾病（CNS）、纤维化（Fibrosis）、免疫与炎症（Immunology & Inflammation）以及眼科疾病（Ocular diseases）等。 Zhavoronkov 创办了 ARD（衰老研究与药物发现大会，最初在巴塞尔，后迁至哥本哈根，本次移师至哈佛大学 David Rubenstein Treehouse 举办）。在往届会议中，各大制药巨头对抗衰老的态度发生了里程碑式的转变： - **礼来（Eli Lilly）的探索**：礼来基因医学负责人 Andrew Adams 在大会上正式抛出议题：“GLP-1 药物是否属于长寿疗法？”，并出示了大量真实世界数据，证明 GLP-1 受体激动剂能在低剂量下预防多种疾病并显著降低多种与衰老相关疾病的发病率。 - **诺和诺德（Novo Nordisk）的定调**：诺和诺德的研发领军人物 Lotte Bjerre Knudsen 随后展示了重要证据，直指司美格鲁肽（Semaglutide）是目前临床证据最充分的长寿药物。 - **全身性生理与心理调节**：Zhavoronkov 分享了自己服用 GLP-1 药物的切身体验，指出该药物不仅能调节全身代谢，还能直接作用于中枢神经以改变心理认知，“消灭了我脑海中所有驱使我在深夜暴饮暴食的恶魔”，并显著减轻了心理压力。GLP-1 的成功为后续双重用途抗衰老药物铺平了临床与商业化路径。 --- ### [11:35](https://www.youtube.com/watch?v=1_ibDuLTCfA&t=695s) AI 赋能的靶点发现与 TNIK 候选药破局特发性肺纤维化 2022 年，英矽智能发表了总结多年 AI 研究成果的学术论文，系统展示了如何将潜在药物靶点按照新颖性、成药性（Druggability）、商业可行性以及与衰老标志物的关联度进行全面映射。 回溯至 2019 年，英矽智能完成了 3700 万美元的 A 轮融资，开始了一场从统计学上看“成功率低于 1%”的冒险之旅——利用纯 AI 驱动，从零发现一个全新的抗衰老兼抗疾病靶点，并设计出全新分子推进到临床阶段。团队甚至为此同步拍摄了纪录片。 - **多通路全新靶点 TNIK**：通过生成式 AI 算法，团队识别出了关键蛋白靶点 TNIK（Traf2- and Nck-interacting kinase）。该靶点能够同时调控 TGF-β、TNF-α、NF-κB 和 Wnt 四条核心信号通路，在多种纤维化与炎性疾病中扮演枢纽角色。 - **锁定首发适应症——特发性肺纤维化（IPF）**：IPF 是一种典型的衰老驱动型致命疾病，平均发病年龄在 65 岁左右。健康人群在 65 岁后每年肺活量（用力肺活量，FVC）自然衰退约 30 至 40 毫升；而 IPF 患者在未经治疗的情况下，每年 FVC 损失高达 400 毫升，且现有标准疗法效果非常有限。 --- ### [15:09](https://www.youtube.com/watch?v=1_ibDuLTCfA&t=909s) 从 79 个分子到临床逆转：TNIK 抑制剂的突破与全自动生物机器人实验室 2024 年 3 月，英矽智能在《Nature Biotechnology》上发表了标志性论文，历时近一年的审稿，完整复盘了从 AI 靶点发现到分子生成设计的全流程。在小分子化学设计阶段，AI 系统展现出了极高的命中率： - **极致的合成效率**：整个项目仅合成了 79 个分子便锁定了最终候选物（传统药企通常需要合成并测试数千个分子），而药物分子的单次初筛合成成本通常在 2500 至 5000 美元之间，少合成化合物直接大幅压缩了研发成本与周期。 - **全自动化机器人实验室验证**：为了支撑高通量生物学实验，Zhavoronkov 先后在苏州和上海建立了大规模全自动化生物机器人实验室。实验证实，该 TNIK 抑制剂具备显著的衰老调控（Senomorphic）特性，在逆转药物诱导型和复制型细胞衰老方面，表现均优于二甲双胍（Metformin）和雷帕霉素（Rapamycin）。 - **2 期临床试验数据突破**：在中国开展的为期 12 周、纳入 71 名患者的双盲安慰剂对照 2 期临床试验结果发表于《Nature Medicine》。数据显示，在高剂量组中，患者的用力肺活量（FVC）不仅停止了下降趋势，甚至呈现出逆转与改善的早期积极信号，同时表现出极佳的安全性。 - **衰老生物钟检测**：团队在试验期间多次采集了患者血液的全面蛋白质组学（Proteomics）表达数据，并联合外部多个实验室运用 6 种不同的衰老生物钟进行评估，多项指标初现生物学年龄逆转的迹象，相关成果正在整理投稿中。 --- ### [18:50](https://www.youtube.com/watch?v=1_ibDuLTCfA&t=1130s) 终极愿景：“生命模型”与 2028 年浦东长寿智慧社区 在演讲尾声，Zhavoronkov 展望了生成式 AI 与长寿科技融合的终极形态。早在 2021 年，他就与计算机科学家李开复博士（Dr. Kai-Fu Lee）以及知名长寿医学专家 Evelyne Bischof 博士共同发表论文，提出了训练“生命模型”（Life Models）的概念——利用生成式 AI 学习人类从出生到死亡全周期的多模态生物学与健康数据。 基于这一理念，英矽智能正在上海浦东推进一个定于 2028 年第三季度交付的大型长寿智慧社区项目： - **产城融合空间**：该社区整合了高品质住宅、商业综合体、餐厅、五星级酒店以及大量入驻的顶尖生物科技企业，构成一体化生命科学创新中心。 - **数据驱动的主动健康防御**：居民的日常健康数据可在获得授权的前提下直接用于现场生物医学研究；智慧配套医院将结合全周期数据进行实时健康干预，目标是“不让任何人因为愚蠢且可预防的原因死去”。 - **沉浸式科研氛围**：社区整体采用极具未来感的建筑设计，园区内将有各类机器人与人形机器人穿梭作业，旨在营造浓厚的科技与科研氛围，“让孩子们不再沉迷于刷 TikTok，而是从小立志投身生物医药科研”。Zhavoronkov 欢迎全球长寿领域的科研人员与产业伙伴在 2028 年汇聚于此。 ## Notable quotes > [00:04](https://www.youtube.com/watch?v=1_ibDuLTCfA&t=4s) 整个医药工业至今未能研发出哪怕一种能够在临床上明确证实“可为健康人额外延长一年寿命”或实现生物学年龄逆转的药物。 — Alex Zhavoronkov > [02:00](https://www.youtube.com/watch?v=1_ibDuLTCfA&t=120s) 从项目立项到确定临床前候选化合物这一阶段，是 AI 能够赋予药企最大“超能力”的无监管自由探索区间。 — Alex Zhavoronkov > [05:18](https://www.youtube.com/watch?v=1_ibDuLTCfA&t=318s) 单纯依托中国的这套高效研发体系，就能在临床前阶段省下约两年的时间；若再叠加端到端 AI 技术的赋能，整个药物发现与临床前开发进程可提速近 4 倍。 — Alex Zhavoronkov > [05:18](https://www.youtube.com/watch?v=1_ibDuLTCfA&t=318s) 一旦药物正式上市并被证实兼具延缓衰老的效果，就相当于“免费完成了抗衰老药物的开发”。 — Alex Zhavoronkov > [08:44](https://www.youtube.com/watch?v=1_ibDuLTCfA&t=524s) GLP-1 药物是否属于长寿疗法？ — Andrew Adams > [08:44](https://www.youtube.com/watch?v=1_ibDuLTCfA&t=524s) 消灭了我脑海中所有驱使我在深夜暴饮暴食的恶魔 — Alex Zhavoronkov > [11:35](https://www.youtube.com/watch?v=1_ibDuLTCfA&t=695s) 成功率低于 1% — Alex Zhavoronkov > [15:09](https://www.youtube.com/watch?v=1_ibDuLTCfA&t=909s) 整个项目仅合成了 79 个分子便锁定了最终候选物 — Alex Zhavoronkov > [15:09](https://www.youtube.com/watch?v=1_ibDuLTCfA&t=909s) 在高剂量组中，患者的用力肺活量（FVC）不仅停止了下降趋势，甚至呈现出逆转与改善的早期积极信号，同时表现出极佳的安全性。 — Alex Zhavoronkov > [18:50](https://www.youtube.com/watch?v=1_ibDuLTCfA&t=1130s) 不让任何人因为愚蠢且可预防的原因死去 — Alex Zhavoronkov > [18:50](https://www.youtube.com/watch?v=1_ibDuLTCfA&t=1130s) 让孩子们不再沉迷于刷 TikTok，而是从小立志投身生物医药科研 — Alex Zhavoronkov ## Mentioned Alex Zhavoronkov, 英矽智能 (Insilico Medicine), TNIK, 特发性肺纤维化 (IPF), GLP-1, 司美格鲁肽 (Semaglutide), 二甲双胍 (Metformin), 雷帕霉素 (Rapamycin), 礼来 (Eli Lilly), Andrew Adams, 诺和诺德 (Novo Nordisk), Lotte Bjerre Knudsen, ARD (衰老研究与药物发现大会), Nature Biotechnology, Nature Medicine, 李开复 (Kai-Fu Lee), Evelyne Bischof, Science MMAI Gym ## Video description Alex Zhavoronkov - Founder and CEO, Insilico Medicine How will generative AI and automation transform the pharmaceutical industry? In this presentation, Dr. Alex Zhavoronkov, CEO of Insilico Medicine, will detail how generative AI, automated platforms, and AI agents will accelerate the discovery of novel biological targets and the design of new molecular structures. He will explain how reinforcement learning and end-to-end automation will enable drug candidates to move into human clinical trials in record time while significantly reducing the cost and duration of the preclinical phase. The talk will also show how these AI capabilities will extend beyond traditional drug discovery into aging research, agricultural chemistry, and environmental sustainability. He will also introduce Science MMAI Gym, a domain-specific training environment designed to transform any causal or frontier Large Language Model (LLM) into a high-performance engine for real-world drug discovery and development tasks.Overall, the session will present a benchmark for the current and emerging capabilities of end-to-end AI in biotechnology and its potential to extend human longevity. [Source](https://www.youtube.com/watch?v=1_ibDuLTCfA)

Source quotes

[0:04] Alex Zhavoronkov: 整个医药工业至今未能研发出哪怕一种能够在临床上明确证实“可为健康人额外延长一年寿命”或实现生物学年龄逆转的药物。 [2:00] Alex Zhavoronkov: 从项目立项到确定临床前候选化合物这一阶段，是 AI 能够赋予药企最大“超能力”的无监管自由探索区间。 [5:18] Alex Zhavoronkov: 单纯依托中国的这套高效研发体系，就能在临床前阶段省下约两年的时间；若再叠加端到端 AI 技术的赋能，整个药物发现与临床前开发进程可提速近 4 倍。 [5:18] Alex Zhavoronkov: 一旦药物正式上市并被证实兼具延缓衰老的效果，就相当于“免费完成了抗衰老药物的开发”。 [8:44] Andrew Adams: GLP-1 药物是否属于长寿疗法？ [8:44] Alex Zhavoronkov: 消灭了我脑海中所有驱使我在深夜暴饮暴食的恶魔 [11:35] Alex Zhavoronkov: 成功率低于 1% [15:09] Alex Zhavoronkov: 整个项目仅合成了 79 个分子便锁定了最终候选物 [15:09] Alex Zhavoronkov: 在高剂量组中，患者的用力肺活量（FVC）不仅停止了下降趋势，甚至呈现出逆转与改善的早期积极信号，同时表现出极佳的安全性。 [18:50] Alex Zhavoronkov: 不让任何人因为愚蠢且可预防的原因死去 [18:50] Alex Zhavoronkov: 让孩子们不再沉迷于刷 TikTok，而是从小立志投身生物医药科研

Source quotes

[0:04] Alex Zhavoronkov: If you don't know how to do it, you can ask your favorite LLM. [5:18] Alex Zhavoronkov: You kind of do free aging research by aligning yourself with a traditional pharmaceutical drug discovery and development paradigm. [8:44] Alex Zhavoronkov: kills all the demons in my head that want me to eat in the evening [18:50] Alex Zhavoronkov: the hospital will not let you die of something stupid.

Original transcript

[0:04] Hi everybody. Uh my name is Alex Shankov. Uh and  I'm and CEO of Encilico Medicine. And uh today I [0:11] hope to inspire some of you to go into aging and  longevity research with AI because nowadays the [0:20] barrier to entry is um much lower than it used to  be. Uh if you don't know how to do it, you can ask [0:26] your favorite LLM. uh and uh uh at Encilica we've  developed pathways uh and methods uh for how to [0:36] uh get some of the dualpurpose uh disease and  uh aging potentially longevity therapeutics [0:44] uh into the uh clinical realm and I'm going to  talk about uh how we how we do that. Uh so full [0:51] disclosure I kind of don't buy sell any securities  based on what I say because we are public company. [0:57] Uh but most importantly don't take any drugs based  on what I say because I'm not a medical doctor [1:02] and everything we're going to be talking about is  highly investigational. Uh so my biggest problem [1:08] with life in general uh is that over the past  150 years uh of pharmaceutical research uh our [1:17] industry has not has failed to produce a single  drug that can give give you one year of life if [1:23] you are already healthy optimizing uh and doing  regular diagnostics. So even if you are very rich [1:31] uh and you are very intelligent and you know uh  about the drugs uh whatever you take right now [1:39] uh has never been clinically proven to work in  context of a human clinical trial showing age [1:44] reversal uh or significant lifespan benefit unless  you have a disease or some medical condition. So [1:52] how do we discover drugs that target aging? It's  very very difficult. So to do that we actually [2:00] tried to so I'm like religiously dedicated to this  topic and I was rich before in Silicon um and I'm [2:07] poor because uh we kind of uh were running out  of money several times uh but now we are kind of [2:14] rich again so to speak um and uh trying to build  the sustainable business model to ensure that we [2:22] can have the resources um in order to develop  and have the time to develop those longevity [2:28] therapeutics. Uh so we're truly global. Uh Boston,  Montreal, Abu Dhabi, and uh um uh greater China. [2:37] Uh and uh um one of the ways to kind of create  a sustainable business model in longevity [2:45] therapeutics is to just be very very good at um  the pharmaceutical industry kind of service and [2:52] uh delivering constantly highquality drugs to  the pharmaceutical industry to build to to to [2:58] also build the capabilities in drug discovery.  uh and uh uh the most impactful area where [3:07] um AI gives you significant advantage is this  wonderful time from the time you start the program [3:15] to PCC pre-clinical candidate or we refer we also  refer to it as developmental candidate. Uh so for [3:23] those of you who are unfamiliar with how pharma  works uh this is a very important slide because it [3:28] shows you uh the different stages drug discovery  and development uh from target identification. So [3:36] the time where you identify this kind of protein  criminal that is driving the disease and the time [3:41] you create cufflings for the for this criminal  either in the form of a small molecule or an [3:48] antibbody or some other therapeutic um you you  you spend a lot of time and also a lot of money. [3:57] So here you actually can see the probability  of success at every stage. Uh how much money [4:02] it costs and how much uh time it takes usually. Uh  and AI gives you superpowers uh in the unregulated [4:12] territory from the time you discover a target to  the time you nominate a developmental candidate. [4:17] Developmental candidate in kind of my book  at Encilico we usually have nonGLP toxicity [4:24] studies in at least two species. Usually you  would do uh rat and dog or u uh or or mouse rat [4:32] and monkey uh to prove that the drug is safe. Uh  then you nominate a developmental candidate and [4:38] then you do the GLP toxicity studies. um that's  also a very regulated area uh and costs you uh [4:45] much longer and even though it's like 28 day GLP  toxicity studies they still take you approximately [4:51] 9 months to a year and after that you are moving  with the speed of traffic once you're on clinical [4:56] trials actually pharmaceutical companies do  clinical trials even better than startups because [5:01] uh they have much better relationship with  the hospitals the principal investigators [5:06] uh and the regulators um so in the ideal world as  a drug discovery company in AI, you should not go [5:13] into human clinical trials yourself. You should  license those drugs to pharma uh in the ideal [5:18] world. So there is a new reality that is emerging  in our field. So one is um China has become a [5:28] really great platform to discover therapeutics  because of the integrated infrastructure uh very [5:34] abundant highly talented labor uh that is working  very hard also favorable policies. Uh if you apply [5:44] uh AI to the China platform you can accelerate  drug discovery approximately four times. [5:52] So just without AI you actually save two years  of your preclinical development um by utilizing [5:59] this just very efficient um uh system and at  encilico we were not kind of born in China but [6:06] we realized that most of the competition is there  and most of those competitors are actually not [6:12] uh AI companies they are classical pharmaceutical  companies there uh that are um that superpowers [6:20] in preclinical development. So the best way to  compete is actually go there and compete there. So [6:26] uh we decided to compete in China with China.  Uh and uh we compete there on novelty. Uh but [6:34] local pharmaceutical companies actually  deliver enormous speed. So you can speed [6:40] up your program quite dramatically by just  being there. And uh one other thing that we [6:46] uh where we are different from anybody else uh  in China and also in um the rest of the world [6:52] is that when we select therapeutic targets uh for  clinical development or or preclinical discovery [7:00] uh we try to identify targets that are also  implicated in aging and disease at the same [7:05] time. So the idea is to get the drug approved  for a disease and then if it actually works in [7:10] aging well it's free. So you kind of uh  uh do free aging research by uh aligning [7:17] yourself with a traditional uh pharmaceutical drug  discovery and development paradigm. So we try to [7:23] identify the protein targets that are supported  by aging clocks uh coming from human data and also [7:29] animal data and also are supported by hallmarks  of aging. So uh same processes that drive many [7:36] diseases are implicated on aging as well. Aging  is the process that all of you have uh that I [7:41] have that's my greatest enemy and my greatest  problem in life I think uh maybe your greatest [7:48] problem in life as well because it's going to  take everything away from you at some point [7:51] in time. Uh so we need to have those dualpurpose  therapeutics that may have a chance to uh reverse [7:58] your biological age and give you a few years of  life. So one uh approach that uh we've developed [8:04] at Encilico is that we try to identify drugs that  um may be what we call pipeline on a product. So [8:11] one drug that may work on many diseases at the  same time. You know that aspirin you would take [8:16] for many different diseases. I you can take it  for you know a headache or for fever for a blood [8:23] thinning many many other things. So here the same  idea. So you identify a novel target that may be [8:28] implicated on one disease, try to get it approved  and then expand the indication beyond um beyond [8:35] your original disease and hopefully at some point  in time develop it for aging. Uh so is there is [8:44] it possible to actually develop therapeutics for  aging in a traditional pharma paradigm? Turns out [8:51] there is. So in pharma longevity biotechnology is  cardio metabolism. Uh it is basically drugs that [9:01] help you regulate your metabolism things  like GLP-1 uh CNS. So drugs that target [9:07] neurodeenerative diseases, fibrosis, imunology,  inflammation and actually many other disease areas [9:13] are also like ocular diseases are also related  to uh to aging quite a bit. [snorts] And uh [9:20] uh I founded a conference called ARD, Aging  Research and Drug Discovery. Originally in Basil, [9:25] then it moved to Copenhagen. This year it's going  to be uh in Boston um at the David Rubenstein's [9:30] Treehouse at Harvard 1st of October, 3rd of  October. So please mark your calendars if you [9:35] like. But uh last year uh Andrew Adams uh the  head of um genetic medicine at La Lilly Lilly [9:43] in my opinion is the most impactful longevity  company in the world. Um because how many of [9:49] you are on GLP1s here? Um well see it may means  that there is a lot of potential for this drug [9:56] in in this room because I am on this drug. Um and  there is significant evidence showing that GLP1s [10:03] uh prevent at low dose prevent disease uh and uh  may actually improve some biological conditions. [10:11] uh at this event for the first time uh Lily uh  presented a question are uh GLP1's longevity [10:18] therapeutics and provided a lot of evidence that  uh many age related diseases may be slow slowed [10:25] down or uh you can decrease the incidence of  them um with GLP-1 just from real world data [10:31] uh and it's like uh third party evidence and then  the next day loten who might be might get a Nobel [10:40] Prize for the discover discovery of JLP1s at  Nova Nordisk presented this wonderful slide [10:45] showing that semiglutide is a proven longevity  medicine is the most proven longevity medicine. [10:51] Uh why? Because again it impacts pretty much every  area of your body modulating metabolism including [11:00] your psychology your desire to eat or do some  bad things. um like that's the reason I take it [11:08] myself because it kind of kills all the demons  in my head that want me to eat in the evening. [11:14] Kind of reduces stress. So you can actually now  see that GLP1s created a pathway for longevity [11:22] therapeutics and now we can just scale that in  many other diseases and uh that's by the way the [11:28] conference if you are interested um this year  it's going to be truly epic. So the idea is to [11:35] utilize AI to identify therapeutic targets uh  those proteins that are um kind of causing disease [11:42] and aging at the same time uh by uh looking at the  overlap and we've published multiple papers on how [11:49] to do that. So one of the papers we've published  2022 describing several years of our research [11:56] uh utilizing AI uh where we showed how to map  therapeutic targets into hallmarks of aging. So [12:03] now we know several core biological processes that  drive aging. They're called hallmarks of aging. [12:09] uh and now we can map some of the biological  targets by novelty by drugability by their [12:16] um commercial tractability but also by their  implication in the hallmarks of aging. So in 200 [12:25] uh I'll skip a few slides. In 2019 uh we decided  to set on a journey. So we got a enough money. So [12:33] that was our first big uh venture round 37 million  from expert by technology investors. Um we decided [12:42] to set on a journey to try to identify a novel  therapeutic target um novel molecule for aging [12:49] and disease at the same time. uh and uh uh the  probability of something like that happening is [12:55] like less than 1%. Uh just statistically. So for  a novel target very often they don't reach even [13:01] clinical trials. But we decided to try even start  shooting documentary movie about that. So how do [13:07] you do that? So first you need to um you need  to well get enough funding. Uh then you need to [13:16] um discover a target that your team will believe  in. uh and uh you can also get expert drug hunters [13:24] to align with you in your vision. Uh and then  you need to uh develop a molecule that is [13:29] safe and effective at a preclinical level for a  specific disease. Uh and then run clinical trials [13:36] uh all the way to efficacy studies phase two in  real patients uh gathering the data you need to [13:42] prove that the drug may have longevity benefits.  That is mission impossible. So before that we [13:48] published multiple papers how we're going to do  that over time uh in bionformatics in chemistry in [13:55] large language models in longevity research u and  demonstrated that we've identified a new target uh [14:03] called tenic that is modulating three pathways at  the same time TGF beta TNF alpha NFAPA B um and W [14:12] for those of you who know the pathways implicated  this target into multiple diseases is uh and then [14:18] we actually asked the question so which disease  do we go after in order to prove the point and we [14:23] decided to go after idiopathic pulmonary fibrosis  because that's the disease of aging average age of [14:29] onset is about 65 uh and uh the outcome measure  that you would use to measure the efficacy of [14:36] the drug is called force val capacity. So how much  air can you breathe out of your lung? And after 65 [14:43] uh even for normal people, you start losing about  like 30 to 40 milliliters of lung capacity every [14:47] year year anyway. Uh in IPF you lose like 400  milliliters without treatment. Um it's a pretty [14:54] dramatic disease. Uh and um currently there  is no good standard of care. Uh so we chose [15:00] um IPF as our first indication for for our tenic  program. uh and uh um I'll skip a few slides. I [15:09] also explained how we're going to do it uh in  an academic paper uh basically going from clock [15:15] to clock. Um so if you want to uh kind of learn  how to develop drugs for aging um in the proper [15:23] pharmacological context, this is a good uh paper  to read. Uh and then uh we actually presented a [15:30] journey. So this is probably the most important  paper to date we published in nature biotechnology [15:35] 2024 in March. Uh it took almost a year to  publish. Uh in that particular paper we showed the [15:42] history of the discovery of the target. Uh all the  way from uh the AI that identified the therapeutic [15:49] target and implicated it on aging all the way to  small molecule chemistry design. We synthesized [15:54] only 79 molecules for this program. Uh, usually  pharmaceutical companies synthesize thousands. And [16:00] just to FYI, like your first date with a molecule  to kind of say hello cost you about $2 and a half [16:06] to $5,000. So the less you synthesize the um the  better. Uh, and we we synthesized only 79. Uh, [16:14] some of them worked very very well. Most of them  worked very well. uh we tested it in many many [16:20] pre-clinical experiments and many mouse models uh  in uh dog and monkey uh and then took it all the [16:27] way to phase one human clinical trials completed  and all that on one paper and then we started [16:32] two phase 2s one on the US one in China and while  we were in phase two we also showed so I built a [16:39] massive uh robotics facility that allows you to do  biological experimentation um in a seamless fully [16:45] automated way in originally in SUJA Now I built  another one in Shanghai um and showed in that [16:52] paper that this drug has cenomorphic properties.  It can reverse um uh cellular scinessence by by [17:01] by certain extent in both drug induced and um and  replicative scinessence better than morphine and [17:10] better than sorry better than um uh metformin and  better than ropomyosin. uh and uh uh in the in the [17:19] um clinical trial. So the China trial completed  very quickly. Uh it took approximately one and a [17:26] half years to complete the trial. We published the  results in nature medicine. It's a proper placebo [17:32] controlled blinded double blinded trial. Uh and  we showed uh this is extended uh figure uh 2A [17:41] um improvement in a very promising trend uh in  force 12 capacity. Uh so not just slowing it down [17:50] uh slowing down the decline but actually reversing  it at high dose. Uh the results are early [17:55] uh and it's only a 12week trial study 71 uh  patients uh but it's very promising. So we [18:02] showed promising safety and efficacy and also uh  demonstrated that we collected proteomics data [18:10] uh from patients. So looked at all the netitality  of the protein expression in their blood several [18:16] times during the trial. Uh and uh now we have  one more paper to publish uh hopefully. So I can [18:24] I don't want to kind of make any claims yet. We  usually don't do it until the paper is published. [18:29] uh showing that we have applied aging clocks in  the context of a clinical trial hopefully showing [18:35] uh some age reversal. Um and now we have uh  six clocks that we used in in the study from [18:43] different labs. Um and uh some of them kind of  showed promising results. I'm almost out of time, [18:50] but I just wanted to show you one project that  we're working on. Uh so imagine we published [18:56] this paper on kind of AI for longevity  research in 2021 with Dr. Dr. Lee Ka who [19:02] you probably know pretty famous um scientist in  AI and one of the most famous longevity doctors [19:10] um Edwin Bishop. Uh the so now you can use  generative AI to collect uh to train life models, [19:18] models of human life that understand you from  the time you are born to the time you die. And [19:24] um uh based on this paper, we set on the project  to develop um uh to develop a community uh where [19:33] you can live, work, do great science and it will  not let you die of something stupid. Uh so it's [19:40] uh launching in 2028 Q3 in Pudong. Um so it has  residential real estate where you can live and uh [19:49] enjoy life. uh commercial real estate, uh shopping  malls, uh restaurants, um five-star hotel and [19:56] also massive number of biotech companies uh in one  place. Uh and your data can be used for biomedical [20:03] research on site. Uh plus uh if you like to give  it to a hospital, the hospital will not let you [20:09] die of something stupid. [snorts] Uh so it's  being built right now. Um, we also want to make [20:15] sure that it's kind of an architectural marvel  so you can walk around it, see the robots. Uh, [20:20] so your kids don't go into like Tik Tok and  they go into, you know, biomedical research. Um, [20:26] and your humanoids too. Um, so targeting 2028 Q3  uh and we published a paper on that. This is the [20:35] current status um, uh, really centrally located.  So if you are planning to you know do really cool [20:42] biomedical research uh in longevity biotechnology  that's a really good place to to be. Uh so thank [20:49] you very much. Uh talk to me if you are interested  in longevity or follow us on LinkedIn. Thank you.