Insilico Medicine has introduced the Virtual Aging Cell (VAC), an artificial intelligence platform that simulates how cells change and age over time, marking a significant shift in computational biology research. Unlike existing virtual cell models that capture cells as static moments, the VAC platform integrates biological age as a core variable and uses multiple AI agents working together to model cellular processes across six biological scales, from molecular interactions to whole organisms.

Why Does Modeling Aging Cells Matter for Drug Discovery?

Most current virtual cell models rely on data collected at single points in time, treating cells as frozen biological units. This approach misses the dynamic processes that actually define living cells, including division, differentiation, aging, and response to environmental changes. Insilico's new platform addresses this limitation by building biological age directly into its computational framework, allowing researchers to simulate how cells evolve and respond to interventions over their lifespan.

The VAC platform uses what Insilico calls a "multi-agent AI architecture," meaning multiple specialized AI systems collaborate to reason across different biological levels simultaneously. This approach enables the platform to model complex interactions between intracellular molecular networks, communication between cells, tissue microenvironments, and individual genetic backgrounds, all while accounting for how these systems change with age.

"Every computational model needs a first principle, and ours is biological age. 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," said Alex Zhavoronkov, Founder, Co-CEO, and Chief Business Officer of Insilico Medicine.

Alex Zhavoronkov, Founder, Co-CEO, and Chief Business Officer of Insilico Medicine

How Did Insilico Build This Technology Over a Decade?

The VAC platform represents the culmination of ten years of research and development. The journey began in 2014 when Insilico first proposed the concept of computational virtual cells at an NVIDIA conference. The company then partnered with BioTime, a regenerative medicine pioneer, on a collaboration called Embryonic.AI, which used deep neural networks to track how cells differentiate and make fate decisions during embryonic development.

In 2017, this partnership produced the first experimentally validated result: AI algorithms identified COX7A1 as a key factor controlling the transition from embryonic to fetal development. This success demonstrated that artificial intelligence could precisely identify the molecular "switches" governing cell differentiation and reprogramming, laying the biological foundation for virtual cell simulation.

As large language models and transformer technology advanced, Insilico shifted its approach. The company developed the PreciousGPT series, a family of scientific foundation models specifically designed for aging research:

• Precious1GPT (2023): An aging clock based on multimodal transformers that integrated DNA methylation and gene expression data to predict biological age and distinguish disease from healthy samples. The model identified dual-purpose targets like APLNR and IL23R with both anti-aging and anti-disease potential.
• Precious2GPT (2024): Expanded capabilities from prediction to conditional generation, combining a conditional diffusion model with a multi-omics transformer. It could generate synthetic biological data with specific tissue and age characteristics, with accuracy exceeding baseline models like conditional GANs.
• Precious3GPT (2024): A true multimodal transformer integrating text, tabular data, and knowledge graphs across four species (mouse, rat, monkey, and human) and three biological data types (gene expression, DNA methylation, and proteins). It performed cross-species age prediction, target discovery, and compound sensitivity prediction in a unified framework.

What Can Researchers Do With Virtual Aging Cells?

The VAC platform is designed to support several critical research applications in drug discovery and aging science. By simulating how cells age and respond to interventions, researchers can predict drug responses more accurately, understand disease mechanisms at the cellular level, identify new therapeutic targets, and explore cellular reprogramming and fate intervention strategies.

The platform's ability to model cells across multiple biological scales simultaneously is particularly powerful. Rather than studying molecular interactions in isolation, researchers can now simulate how changes at the molecular level propagate through intracellular networks, affect communication between cells, reshape tissue environments, and ultimately influence organ and organism-level outcomes. This systems-level perspective could accelerate the discovery of drugs that work more effectively and have fewer unintended side effects.

Insilico has launched a dedicated Virtual Aging Cell webpage and is previewing the multi-agent VAC generation platform for collaboration with pharmaceutical companies and research institutions. The company is positioning the technology as a frontier tool for modeling both healthy and diseased states across virtual cells, virtual systems, and virtual populations.

How Does This Fit Into the Broader AI Research Landscape?

Virtual cells were recognized as one of seven technology breakthroughs to watch in 2025 by the journal Nature, reflecting growing interest in computational biology across the research community. However, most existing virtual cell models have been limited by their inability to capture temporal dynamics and multi-scale interactions. Insilico's integration of biological age as a core conditional variable represents a conceptual advance that addresses a fundamental gap in how computational models represent living systems.

The VAC platform also reflects a broader trend in AI research toward more specialized, domain-focused models. Rather than relying on general-purpose large language models, Insilico built scientific foundation models trained specifically on aging biology and multi-omics data. This specialization allows the models to capture domain-specific knowledge and relationships that general models might miss.

The research underlying the VAC platform has been published in peer-reviewed journals including Aging, npj Aging (a Nature portfolio journal), and bioRxiv preprints, with collaborations involving Harvard Medical School. The open-sourcing of Precious3GPT on platforms like Hugging Face and GitHub reflects a commitment to making these tools accessible to the broader research community.