Insilico Medicine Launches Industry's First Multi-Agent Virtual Aging Cell Platform Centered on Biological Age
核心洞察
Insilico Medicine (搜索) announced the launch of its Virtual Aging Cell (VAC) Webpage and a preview of its Multi-Agent driven VAC platform, the industry's first virtual cell platform built around biological age as a core condition.
The VAC platform uses a multi-agent AI architecture to perform collaborative reasoning across six biological scales, aiming to dynamically simulate cellular differentiation, reprogramming, and aging for target discovery and drug development.
The launch culminates a decade of research spanning the PreciousGPT foundation model series, which identified dual-purpose targets such as APLNR (搜索) and IL23R (搜索) with anti-aging and anti-age-related-disease potential.
Insilico Medicine (搜索) ("Insilico", HKEX: 3696), a clinical-stage drug discovery company powered by generative artificial intelligence (AI), announced the launch of its Virtual Aging Cell (VAC) Webpage alongside the preview of its Multi-Agent driven VAC platform. The platform represents the industry's first multi-agent Virtual Aging Cell system built around biological age as a core conditional variable, addressing a recognized frontier in computational biology.
Virtual cells are digital models that leverage AI and mathematical modeling to computationally simulate cellular biology and its underlying dynamic laws, offering capabilities to predict drug responses, elucidate disease mechanisms, and support target discovery. Virtual cells were named one of the seven technology breakthroughs to watch in 2025 by Nature.
Addressing the Limits of Static Cell Models
Most existing virtual cell models rely predominantly on data collected at isolated time points, capturing cells as static snapshots. Their capacity to simulate dynamic temporal processes—such as cellular differentiation, reprogramming, and aging—as well as multi-scale interactions across biological hierarchies, remains markedly constrained. Living cells constantly undergo dynamic processes including division, differentiation, aging, and environmental response, with their state shaped by intracellular molecular networks, intercellular communications, tissue microenvironments, and individual background parameters.
The VAC concept directly addresses this frontier. Built around biological age as a core conditional variable, the platform utilizes a multi-agent AI architecture to perform collaborative reasoning across six biological scales—molecular, intracellular, intercellular, tissue, organ, and organism/population. It aims to dynamically simulate cellular differentiation, reprogramming, and aging, offering computational support for target identification, cellular fate intervention, novel drug discovery, and geroscience research.
"Every computational model needs a first principle, and ours is biological age," said Alex Zhavoronkov, PhD, Founder, Co-CEO, and Chief Business Officer of Insilico Medicine (搜索). "This has been a decade-long scientific journey, from raising an industry-defining question at NVIDIA GTC in 2014, to teaching AI the language of aging biology with the PreciousGPT series, and now reimagining the virtual cell through agentic AI swarms."
A Decade of Foundation: From Virtual Cell Concept to PreciousGPT
Insilico's exploration in virtual cells and geroscience spans a decade. At the 2014 NVIDIA GTC conference, the company first articulated its vision for computational virtual cells, virtual organs, and digital populations, posing the question "Can NVIDIA help solve aging?" The company subsequently partnered with BioTime on a collaboration called Embryonic.AI, which used Deep Neural Network Ensembles to track and modulate cell differentiation trajectories during the Embryonic-to-Fetal Transition (EFT). In 2017, the teams published findings in Oncotarget confirming that AI algorithms identified COX7A1 (搜索) as a key transition factor—Insilico's first experimentally validated AI target discovery result.
In May 2022, Dr. Zhavoronkov proposed applying multimodal Transformers to aging research at the Gordon Research Conference (GRC). This led to the PreciousGPT series of scientific foundation models. In 2023, Insilico released Precious1GPT, an aging clock based on multimodal Transformers and transfer learning that integrated DNA methylation and transcriptomic data to predict biological age. Using this model, the team identified dual-purpose targets such as APLNR (搜索) and IL23R (搜索) with both anti-aging and anti-age-related-disease potential, published in Aging.
Released in 2024, Precious2GPT fused a Conditional Diffusion model (CDiffusion) with a decoder-only Multi-Omics Pre-trained Transformer (MoPT), capable of generating high-quality synthetic multi-omics data carrying specific tissue and age characteristics, published in npj Aging. That same year, Insilico collaborated with the Vadim N. Gladyshev Lab at Harvard Medical School to release Precious3GPT, a multimodal Transformer integrating text, tabular data, and knowledge graph representations across four species and three omics modalities, open-sourced on Hugging Face and GitHub.
Agentic AI Architecture: From Molecular Simulation to Organism Evolution
The VAC platform introduces a Multi-Agent Swarms architecture that grants genuine intelligent reasoning capabilities to simulate dynamic living systems, redefining the computational biology paradigm across three core dimensions.
First, biological time serves as a core condition. The multi-agent architecture treats biological age as the core conditional variable that penetrates every scale, allowing the agent swarm to dynamically assess differential responses of tissue microenvironments across young, mature, and aged life stages.
Second, the platform deploys six-scale agent swarm collaboration. Master Agents handle global task scheduling and cross-level logic integration, while Specialist Agents execute specialized tasks at each level. Individual agents can independently access databases, omics tools, and literature knowledge graphs, engaging in adaptive negotiation and joint reasoning across six biological scales.
Third, the platform moves from state observation to fate intervention. Unlike traditional virtual cells limited to one-directional state description, the VAC platform is designed to answer "how to reverse and reshape cell fate," reasoning in real time about microenvironment changes, downstream pathway cascades, and cross-scale structural remodeling for stem cell reprogramming and systemic anti-aging strategies.
"What distinguishes VAC is not the number of hierarchical levels or agents, but the combination of bidirectional Master Agent coordination and first-class organism context storage," said Vladimir Naumov, PhD, Head of TargetID of Insilico Medicine (搜索). "Among the papers we surveyed, no existing system simultaneously possesses these features: six biological hierarchies, an agent for every biological entity, dual top-down and bottom-up Master Agents, and a shared organism context that prioritizes age as a variable."
"Biology unfolds across time and interconnected levels, yet many computational models examine only one layer or moment in isolation," said Petrina Kamya, PhD, Vice President, Global Head of AI Platforms and President of Insilico Medicine (搜索) Canada. "VAC is designed to bring those dimensions together, combining biological age with multi-agent reasoning to explore how changes in a cell's age affect tissues, organs, and the broader organism."
Clinical Momentum and Future Horizons
Insilico reports that while traditional early-stage drug discovery typically spans 2.5 to 4 years, the company consistently reaches developmental candidate (DC) nomination in an average of 12 to 18 months, synthesizing and testing only 60 to 200 molecules per program. Since 2021, the company has nominated 33 PCCs, 13 of which have received IND approvals or clearances, with 9 DC nominations announced since the beginning of 2026.
Insilico's most advanced program, Rentosertib (搜索), completed a Phase IIa clinical trial establishing proof of concept for safety and efficacy trends, with findings published in Nature Medicine in March 2025. This track record, combined with the proprietary data accumulated through its R&D engine, provides real-world grounding for the VAC platform's dry-lab and wet-lab validation loop and target authenticity.
With the launch of the Virtual Aging Cell Webpage, Insilico has released an initial demo preview video of the VAC platform. The company will continue the conversation at ARDD 2026 (the 13th Aging Research and Drug Discovery Conference) in Boston, engaging with academic and industry partners on the next steps in virtual cell platform development.