> ## Content Index
> Fetch the complete content index at: https://www.bushletter.com/llms.txt
> Use this file to discover other available public pages before exploring further.

# Nobel laureate's startup builds the first virtual human cell
- URL: https://www.bushletter.com/genbio-ai-builds-a-virtual-human-cell-to-speed-drug-discovery/
- Published: 2026-08-19T22:00:00.000Z
- Updated: 2026-08-19T22:00:00.000Z
- Description: Drug discovery has a long-standing and expensive flaw. Only about one in 10,000 compounds entering preclinical pipelines ever reaches the clinic, a failure rate that has resisted every wave of laboratory automation for decades.
- Author: Editor
- Tags: Technology, Science, US, GenBio AI

![Zara Kincaid](https://res.cloudinary.com/dz77sb7j1/image/upload/v1774262611/bushletter/authors/zara-kincaid.png)

By **Zara Kincaid** · 2026-08-19

TLDR

GenBio AI has released AIDO Cell, a multiscale AI model that simulates human cell behaviour from DNA through to whole-cell dynamics. Co-founded by Nobel laureate David Baker, the system targets drug discovery's brutal attrition rate. The current release is an early preview supporting only two cell lines.

KEY TAKEAWAYS

01Nobel laureate David Baker co-founded GenBio AI, which released AIDO Cell as a virtual human cell simulator.

02Only one in 10,000 drug compounds entering preclinical pipelines ever reaches the clinic, the company says.

03AIDO Cell is stateful: knock out a gene and the effects cascade across every biological scale.

04Current support covers just two immortalised cell lines, K562 and HepG2, limiting immediate applications.

05A Nature Medicine Perspective by Xing, Eran Segal and Le Song maps a three-stage roadmap for AI-driven digital organisms.

## A city, not a house

Drug discovery has a long-standing and expensive flaw. Only about one in 10,000 compounds entering preclinical pipelines ever reaches the clinic, a failure rate that has resisted every wave of laboratory automation for decades.[\[1\]](https://www.statnews.com/2026/08/18/david-baker-genbio-ai-new-virtual-model-unveiled-aido-cell/?ref=bushletter.com) Researchers can model a molecule well enough, but modelling how a living cell responds to that molecule across every biological layer at once has remained out of reach.

Palo Alto startup GenBio AI is making a direct claim on that problem. On 18 August 2026 it released AIDO Cell, which it describes as the first virtual cell world model able to simulate human cell behaviour across the full biological hierarchy, from DNA and RNA through proteins to whole-cell dynamics, in a natural state and in response to drugs.[\[1\]](https://www.statnews.com/2026/08/18/david-baker-genbio-ai-new-virtual-model-unveiled-aido-cell/?ref=bushletter.com)

## Who built it and why it matters

GenBio AI was co-founded by 2024 Nobel laureate in Chemistry David Baker and AI scientist Eric Xing.[\[2\]](https://genbio.ai/home-2/?ref=bushletter.com) Baker is best known for his work on protein structure prediction and design, a field that AlphaFold supercharged. His presence on the founding team signals something about the project's ambition: AIDO Cell is meant to be to the whole cell what AlphaFold was to the single protein.

Baker's own framing makes the scale of the claim plain. "With AlphaFold, you're predicting the structure of one house or learning about one person in that city, whereas in AIDO Cell, you're sort of predicting how that whole city works," Baker said.[\[1\]](https://www.statnews.com/2026/08/18/david-baker-genbio-ai-new-virtual-model-unveiled-aido-cell/?ref=bushletter.com) The analogy is deliberately architectural: a house can be described in isolation, but a city only makes sense as a system of interactions.

## How the model actually works

AIDO Cell is not a single neural network but a series of single-cell foundation models ranging from 3 million to 650 million parameters, pretrained on 50 million human cells drawn from diverse tissues.[\[3\]](https://genbio.ai/aido-multiscale-foundation-models/?ref=bushletter.com) The architecture is multiscale: it connects what happens at the level of a gene to what eventually emerges at the level of whole-cell behaviour, passing information up and down the biological hierarchy rather than treating each layer as a separate modelling problem.

Crucially, the model is stateful. Knock out a gene and AIDO Cell propagates the consequences across scales, showing how the disruption travels from the genome through protein networks and into observable cellular responses. That cascade behaviour is what distinguishes it from earlier tools that treat genomics, proteomics and cell biology as separate, siloed datasets.

Chief Scientific Officer Ziv Bar-Joseph framed the company's broader goal in terms of unification. "Our goal is not to tackle isolated components, but to build AIDO, a unified model that captures how biological entities interact across every scale," Bar-Joseph said.[\[3\]](https://genbio.ai/aido-multiscale-foundation-models/?ref=bushletter.com) AIDO stands for AI-Driven Digital Organism.

## The drug-discovery case

The commercial argument for virtual cell models follows directly from the attrition numbers. A simulation that reliably predicts how a human cell responds to a compound lets researchers filter out failures before spending years and hundreds of millions of dollars on wet-lab work and clinical trials. Foundation models trained on massive biological datasets promise to break down the silos between DNA, RNA, protein and cellular data, enabling the kind of holistic simulation that could make that filtering possible.[\[3\]](https://genbio.ai/aido-multiscale-foundation-models/?ref=bushletter.com)

GenBio AI has pointed to imatinib, the leukaemia drug, as an illustrative demonstration. AIDO Cell was used to model imatinib's mechanism of action in a relevant cell type, and the result matched what is already known about the drug.[\[1\]](https://www.statnews.com/2026/08/18/david-baker-genbio-ai-new-virtual-model-unveiled-aido-cell/?ref=bushletter.com) Reproducing a known mechanism is a standard first validation step; no new drug target or compound was identified.

## The roadmap behind the release

Alongside the product release, a Nature Medicine Perspective co-authored by Eric Xing, computational biologist Eran Segal and AI researcher Le Song sets out a three-stage approach to building what the authors call an AI-Driven Digital Organism.[\[4\]](https://arxiv.org/abs/2412.06993?ref=bushletter.com) The paper, published in December 2024, provides the conceptual architecture that AIDO Cell is meant to instantiate in its first stage.

The Nature Medicine Perspective maps a three-stage roadmap for progressing from single-cell models through tissue-level simulation to a full digital organism, framing AIDO Cell as an early waypoint rather than a destination.[\[4\]](https://arxiv.org/abs/2412.06993?ref=bushletter.com) The paper predates the AIDO Cell release by eight months, which suggests the scientific framing was developed well before the product was ready to show.

## What is missing from the preview

GenBio AI describes this release as a preview and an early but functional demonstration.[\[1\]](https://www.statnews.com/2026/08/18/david-baker-genbio-ai-new-virtual-model-unveiled-aido-cell/?ref=bushletter.com) That framing is accurate in a specific and limiting sense. AIDO Cell currently supports only two immortalised cell lines, K562 and HepG2, both well-characterised laboratory standbys rather than primary human tissue.[\[1\]](https://www.statnews.com/2026/08/18/david-baker-genbio-ai-new-virtual-model-unveiled-aido-cell/?ref=bushletter.com)

Immortalised cell lines are useful research tools, but they differ in important ways from the primary cells found in actual human tissue. Drug developers targeting diseases that do not manifest in K562 or HepG2 cells will need to wait for expanded coverage before the model becomes directly useful to their pipelines.

A stateful, multiscale simulation that correctly predicts the behaviour of even two well-understood cell lines is a meaningful engineering result, particularly when built on a foundation model pretrained on 50 million human cells from diverse tissues.[\[3\]](https://genbio.ai/aido-multiscale-foundation-models/?ref=bushletter.com) The architecture is credible, the scientific pedigree is serious, and the distance between a two-cell-line preview and a tool that changes pharmaceutical development remains substantial.

SOURCES & CITATIONS

1. [David Baker's GenBio AI unveils AIDO Cell virtual model](https://www.statnews.com/2026/08/18/david-baker-genbio-ai-new-virtual-model-unveiled-aido-cell/?ref=bushletter.com)
2. [GenBio AI company overview](https://genbio.ai/home-2/?ref=bushletter.com)
3. [AIDO Multiscale Foundation Models](https://genbio.ai/aido-multiscale-foundation-models/?ref=bushletter.com)
4. [Nature Medicine Perspective: AI-Driven Digital Organism](https://arxiv.org/abs/2412.06993?ref=bushletter.com)

FREQUENTLY ASKED QUESTIONS

What is AIDO Cell?

AIDO Cell is GenBio AI's virtual human cell model. It uses a series of AI foundation models to simulate cell behaviour across multiple biological scales, from gene-level activity through protein networks to whole-cell responses, and can model how a cell reacts to drugs.

Who founded GenBio AI?

GenBio AI was co-founded by David Baker, who shared the 2024 Nobel Prize in Chemistry for his work on protein design, and AI scientist Eric Xing. Ziv Bar-Joseph serves as Chief Scientific Officer.

What are the current limitations of AIDO Cell?

As of its August 2026 preview release, AIDO Cell supports only two immortalised cell lines, K562 and HepG2\. Its headline demonstration reproduced the known mechanism of imatinib rather than identifying a new drug target.

What is the Nature Medicine Perspective about?

The paper, co-authored by Eric Xing, Eran Segal and Le Song, sets out a three-stage roadmap for building an AI-Driven Digital Organism, moving from single-cell models through tissue simulation to a full digital organism. AIDO Cell represents the first stage of that plan.

![Zara Kincaid](https://res.cloudinary.com/dz77sb7j1/image/upload/v1774262611/bushletter/authors/zara-kincaid.png)

[Zara Kincaid](https://bushletter.com/author/zara-kincaid/?ref=bushletter.com)

Zara Kincaid writes about artificial intelligence and search. Her focus is what happens to businesses when the front page of the internet stops being a list of links and starts being an answer.