THE FUTURE OF PERSONALIZED MEDICINE: LIVE CELLS ON A SILICON CHIP

Mapping all human cells to build a comprehensive 'cell atlas' or roadmap that shows how diseases develop is the aim of a major global project, supported by the Chan Zuckerberg Initiative.

"This effort aligns closely with our company's mission," says Dr. Yoav Nissan-Cohen, co-founder and CEO of TeraCyte. A physicist and a pioneer in Israel's semiconductor industry, Nissan-Cohen co-founded Tower
Semiconductor in 1993. In 2020, together with molecular biologist Dr. Gal Lenz, they launched TeraCyte to unite chip and image processing technologies with molecular biology.

"We're redefining how biological data is captured," says Lenz. "In natural language processing, the revolution began when words were represented mathematically as tokens. We've developed a parallel method for biology: the BioTokenTM, a mathematical representation of a living cell. Imagine being able to ask our system which treatment is best for a specific patient, and getting an answer based not on guesswork, but on real data from that patient's own cells."

The Limitation: Today's Biology is Studied in Snapshots - Through Frozen Moments—While Life Itself is Dynamic
"To understand the impact of this innovation, it's important to know that today's clinical and research tools typically capture static snapshots of cells," he continues. "But life is dynamic. Cells move, respond and change over time. Measuring them at a single point limits insight and hampers the ability to predict patient responses to therapy."

While AI is increasingly used in medicine, this 'snapshot' limitation imposes a ceiling. It reduces raw data into predefined parameters, and in doing so loses much potentially valuable information, limiting our ability to build accurate models for predicting clinical outcomes.

"AI's potential in medicine is limited by the quality of the data it receives," explains Nissan-Cohen. "To truly predict outcomes, we need to understand the biological processes, not just the endpoints. And that's exactly what we're doing."

TeraCyte's technology continuously tracks live cells over time, capturing what happens inside each cell rather than taking a single snapshot. It uses an advanced microscope integrated with a custom silicon chip containing approximately 400,000 micro-wells, each capable of isolating a single cell. Over time, the system records subtle cellular changes such as shape, movement, behavior and function, using high-resolution imaging and AI-driven analysis.

This massive, real-time data stream allows researchers to observe many cells simultaneously, identify cellular subpopulations, track their behavior and develop a comprehensive understanding of cellular behavior across cell populations.

From Observation to Prediction: The BioToken
On the TeraCyte chip, cells grow under tightly controlled conditions — humidity, temperature, CO₂ levels — while the intelligent microscopy system captures their activity. The system distils observed changes, reactions and behaviors into a single data unit: a BioToken, a data card that mathematically encodes each cell's unique profile.

"Instead of a limited 'shopping list' of cell traits, BioTokens allow us to understand what causes differences between cell populations," explains Nissan-Cohen. "AI models built from BioTokens can predict treatment outcomes and uncover hidden biological patterns."

How Are Predictions Made?
Model training begins prior to treatment by analyzing the patient's cells and generating BioTokens. Once treatment outcomes are known, these pre-treatment BioTokens are used to train AI models to predict responses in future patients. "This reverses the traditional biological workflow - instead of first exploring the biology and then forecasting the outcme, we strat with data-driven predictions and work backwards to uncover the underlying biology," says Lenz. "This 'reverse biology' approach is gaining traction in leading research institutes worldwide."

While the AI's decision-making process may not always be fully transparent, with a so-called 'black box' between the BioTokens and the final analysis, the predictions it generates are accurate and clinically valuable. "The reliability of the system outweighs the need to fully decode how the AI reaches its conclusions," he says.

Global Collaboration and Clinical Integration
TeraCyte has successfully completed proof-of-concept and is uncovering novel insights across cell types. As part of ARC, the Sheba Medical Center's prestigious innovation hub, TeraCyte is leading a clinical study to predict whichpatients are likely to experience adverse effects from their cancer treatment, an area where precision is critical. The company is also working with the Weizmann Institute of Science, the Technion Institute of Technology, Harvard Medical School, CalTech and more.

"Our goal is to provide physicians with a compatibility score for each treatment based on the patient's own cells," says Nissan-Cohen. "This will help avoid unnecessary treatments and reduce healthcare costs, possibly even replacing expensive procedures like genetic sequencing."

A Vision for the Future
Dr. Lenz envisages a vast BioToken database to power advanced AI models that will be able to answer real-time medical questions, much as ChatGPT does with text. "Our platform is already a cutting-edge research tool," he says. "In immunology and oncology, existing treatments often succeed in no more than 30 to 40 percent of patients, so it's crucial for both patient outcomes and healthcare economics to avoid therapies that won't work or may even cause harm."

"TeraCyte is on an extraordinary journey," Dr. Nissan-Cohen adds, "with breakthrough technology, global collaborations and partnerships, and a vision that's already taking shape, we believe this is only the beginning. When science, engineering and AI converge, the possibilities are limitless."

To Website>>>
in collaboration with TeraCyte