Vivodyne Builds a Human-Tissue Data Factory for Drug AI
The company’s new facility near San Francisco is designed to produce records of how diseased human tissue responds to treatment. Its early benchmarks are company-reported tissue-model results; clinical value remains the test that matters.
Story brief
3 key pointsVivodyne has opened a San Francisco-area facility that automates experiments on living human tissue, aiming to supply drug-discovery models with cause-and-effect data rather than biological snapshots. Its HIVE system covers 20 tissue types and tracks hundreds of thousands of experiments; the company reports strong results on liver, airway, and bone-marrow benchmarks. Vivodyne says the facility runs at twice U.S....
- 01
HIVE grows 20 human tissue types, autonomously doses them, and monitors responses across hundreds of thousands of experiments.
- 02
Vivodyne reports 94% liver-toxicity prediction, 96% airway-tissue matching, and 100% concordance across 20 chemotherapy drugs.
- 03
The company says its facility operates at twice the throughput of all U.S. animal trials currently being held.
Vivodyne is not claiming that AI has solved drug discovery. Instead, it has opened a facility near San Francisco built to create a kind of evidence the company says drug models lack: records of what happens when living human tissue is exposed to a treatment. HIVE’s tissue-model results are encouraging company-reported benchmarks. They do not yet show that medicines selected with the system fare better in clinical trials.
A robotic lab built for human tissue
Vivodyne calls the site the world’s largest human data center. Its modular HIVE laboratories can grow 20 kinds of human tissue, give them doses autonomously, and monitor the results. Chief executive and co-founder Andrei Georgescu says the facility is already running at twice the throughput of all animal trials being held in the United States.
The immediate pitch is a better preclinical screen: help drugmakers decide which candidates may work before committing to clinical trials, which typically cost tens of millions of dollars. Vivodyne says it is working with multiple major pharmaceutical companies, though it has not named them publicly. The company was spun out of the University of Pennsylvania in 2021 and says it has raised just under $80 million in two Khosla Ventures-led rounds.
The data bet is cause, then effect
Vivodyne’s larger argument is that many drug-discovery models learn from animal tests or snapshots of individual cells and proteins. Those sources can show a biological state without recording the action that produced it. HIVE is designed to generate causal data instead: change something in diseased human tissue, then record the response.
The company says its systems are tracking hundreds of thousands of ongoing experiments in which diseased tissue is exposed to stimuli. Georgescu expects that record to help models identify interventions that produce a desired effect. He sees a particular need in combination therapies, where searching for treatments across multiple biological pathways quickly becomes too large for trial-and-error experimentation.
Clinical decisions are the unresolved proof
The gap between a promising preclinical result and an approved medicine is where the industry still struggles. About 90% of drugs that work in animal testing and enter clinical trials do not receive approval for use in humans. A handful of AI-designed drugs have entered human trials, including one that reached Phase III, but the remaining roadblocks are not necessarily problems AI can solve today.
That makes Vivodyne’s clinical test straightforward, even if it will take time: whether its human-tissue experiments help drugmakers choose candidates that ultimately succeed in people. Until then, HIVE is a scaled bet that better experiments can give drug AI a more useful starting point.
Sources
- techcrunch.comAI isn't close to curing cancer. This startup says it knows what it will take. | TechCrunch