Mapping Cell Fate: The Idea Behind Trailhead’s HD-DoE Platform
Part 2 of a 3-part editorial series on Trailhead Biosystems
By Sirena Meade | Founder and Editor, The Biotech Beacon
Some companies begin with a business plan, while others begin with a product. Trailhead Biosystems began with a question. Originally, the question revolved around how scientists could make better specialized human cells from stem cells. But then it became something far more fundamental. What if biology is too complex to solve the way we have always tried to solve it?
For decades, developmental biologist Jan Jensen was captivated by one of biology’s greatest mysteries: how a single, undifferentiated cell becomes the astonishing diversity of tissues that make up the human body. Every neuron, heart cell, liver cell, and pancreatic beta cell begins from a common origin, yet somehow arrives at a remarkably specific identity. That transformation is known as cell differentiation. It is one of the defining processes of human development as well as being one of the most difficult to reproduce reliably in the laboratory1.
Scientists have spent generations trying to understand it. Their traditional approach has been remarkably productive: form a hypothesis, test a variable, observe the result, refine the hypothesis, and begin again. That method has driven countless discoveries. But as Jensen considered the sheer number of connections governing cell fate, he began to question whether the problem was no longer a shortage of intelligence, effort, or scientific knowledge.
Perhaps the problem was scope. Biology is not controlled by a single switch. Cells respond to many signals at once: growth factors, small molecules, signaling pathways, concentration, timing, oxygen, pH, cell density, neighboring cells, and the physical environment around them. Those influences do not operate independently. One may strengthen, suppress, or completely alter the effects of another.
Trying to study that complexity one variable at a time is like trying to understand an orchestra by listening to a single violin. You will learn something, but you will never hear the whole symphony.
In a 2024 reflection on Trailhead’s origins, Jensen described reaching an uncomfortable conclusion2. To address the problem, he would need to restructure some of his scientific habits in which he had been trained; not because hypothesis-driven science was wrong, but because human intuition might be insufficient to anticipate the vast number of relationships hidden inside a differentiating cell. If those relationships could not be reliably inferred, perhaps they could be measured. That idea became the foundation of what would eventually become Trailhead Biosystems.
In 2013, the State of Ohio backed the concept through the Ohio Third Frontier program. Jensen and an early team spanning Case Western Reserve University and Cleveland Clinic proposed building a system capable of testing biological variables at a scale far beyond what was customary. It was an ambitious bet. The technology did not yet exist as a commercial platform, and the experiments required to demonstrate it were expensive. According to Div Trivedi, one of Trailhead’s earliest team members and current COO, the initial state and industry support totaled approximately $5 million and helped fund the construction and early testing of the system.
The first experimental data arrived in 2014. Instead of testing one biological variable at a time, the team could examine combinations of regulatory factors and measure how cells responded through changes in gene expression. Those data could then be modeled mathematically, revealing relationships among factors that the scientists could take back to the bench and test experimentally. The approach had done what Jensen hoped it could: turn an enormously complex biological process into a design space that could be systematically explored. Once it worked, the project could no longer be understood as merely another academic experiment. It had become the beginning of a platform.
An Unlikely Addition to a Biology Team
Among the people drawn further into the effort was Trivedi, who had arrived in Cleveland from India to attend graduate school at Case Western Reserve University. He was not a developmental biologist by training nor did he come from a molecular biology or stem cell science background. Rather, his training was in engineering, mathematics, operations research, and the application of quantitative tools to complex systems. His coursework crossed between applied mathematics, business and operations. While looking for an internship, he encountered a team working across Case Western Reserve and Cleveland Clinic that needed someone to help model the experimental data.
The scientists, including Jensen, wanted a person who could do that without automatically imposing the same assumptions a biologist might bring to the problem.
“Who could be more unbiased to biology than a mathematician?” Trivedi recalled.
The idea was compelling: combine numbers, mathematics, and experimental biology to better control stem cell differentiation. Trivedi joined as an intern in his early-twenties, excited by the chance to apply his training to a problem unlike anything he had worked on before. He was also, in the earliest stages, given limited biological context to help remove any bias in the calculations.
The scientists selected the genes, factors, and biological questions they wanted to test and designed the experiments with the help of mathematical principles of Design-Of-Experiments (DoE). Once the work had been performed in the laboratory and the data generated, Trivedi modeled the data to make sense of it and identify any patterns that would have been difficult to see by inspection alone. The results went back to the scientists to weigh against what they knew of the biology.
Sometimes a model suggested something that appeared mathematically possible but biologically implausible. That’s when the scientists would push back. At other times, the analysis exposed a relationship the team had not considered. The biologists would return to the laboratory and test it. A feedback loop developed: mathematical prediction followed by biological validation, then new data feeding the next round of modeling. The purpose was never to replace biological expertise but to extend its reach.
Seeing Cell Differentiation as a Process
As Trivedi learned more about the technology, he began to recognize something familiar. The team was studying differentiation not merely as a collection of biological observations, but as a process with inputs, outputs, interactions, and controllable parameters. That perspective came naturally from operations research and engineering.
One day, Trivedi approached Jensen with a question: had he ever heard of Six Sigma? Six Sigma is widely associated with manufacturing and process engineering. Its central premise is that a process can be understood and improved by measuring variation, identifying the parameters that drive outcomes, and building more consistent systems around that knowledge. Jensen was unfamiliar with the term but after reading about it, he returned to Trivedi as the comparison fit well with what they were attempting.
“What we do in manufacturing is control the manufacturing process,” Trivedi explained. “We were taking it one step upstream by trying to understand how we could control the biological parameters that help differentiate a stem cell into a specified cell.”
The embryo’s development from one cell into a complex human body is, in one sense, a process. It unfolds through stages, responds to signals, and moves cells toward different identities. But unlike a conventional manufacturing line, its variables are numerous, interconnected, and often difficult to observe directly. To control such a process, the team first needed to understand which parameters mattered, how strongly they mattered, and how they affected one another. That is where Design-of-Experiments entered the story.
Design-of-Experiments, or DoE, is a mathematical discipline used to study how multiple variables influence an outcome. Rather than changing one factor while holding everything else constant, DoE allows researchers to vary several factors within a carefully designed experimental framework and then use statistics to estimate their individual and combined effects. Industries had already used DoE to better understand manufacturing systems. Trailhead’s founding insight was that the same broad principle could be applied to biology, but only if the dimensionality increased dramatically.
The system would need to examine many potential effectors across many genes, over multiple stages of differentiation, while accounting for interactions among the variables. Thus, the concept of High-Dimensional Design-of-Experiments (HD-DoE®) was born.
Casting a Wider Net
Traditional biological optimization often begins with a hypothesis about one factor. A scientist may increase the concentration of a growth factor and observe the result. Then the scientist changes the timing and perhaps a second signaling molecule is added. Each experiment provides useful information, but the number of possible combinations grows rapidly.
If five variables can each be tested at several concentrations and time points, the experimental space soon becomes too large to explore sequentially. HD-DoE approaches the problem differently. Scientists begin by selecting the biological factors and genes most relevant to the cell type they want to produce. Those choices are informed by existing literature, previous experiments, and subject-matter expertise.
David Llewellyn, CEO of Trailhead explained, “We are not limited to what the literature already tells us. HD-DoE® lets us start right from the beginning and build the map ourselves.”
The platform then “casts a wide net,” as Trivedi described it. Automated systems perform carefully structured experiments that vary combinations of regulatory factors. The team measures how the cells respond, predominantly through gene expression data.
Jennifer Antonchuk, Vice President of Research and Development at Trailhead, said that gene expression gives the team a fast way to compare many conditions. But it is only the beginning.
It provides a snapshot of what the cell is doing at a particular moment. By measuring multiple genes across different combinations of inputs, Trailhead can begin to map how each factor changes the broader biological program. Trivedi offered the example of examining approximately 50 genes within an experiment. Rather than asking how a single factor affects only one gene, the resulting models can show how changing that factor across a defined range may alter many genes simultaneously.
Scientists at Trailhead work in the laboratory to prepare cell samples.
The output is not simply a list of conditions that “worked” or “did not work.” It is a mathematical representation of a biological response. The model can estimate boundaries, interactions, and directional effects: how much a signal matters, where its effects change, which genes rise or fall together, and which combinations may move a differentiating cell toward, or away from, the desired identity. In simpler terms, HD-DoE does not merely test more recipes. It attempts to reveal the landscape underneath them.
From a Wide Landscape to a Specific Path
Stem cell differentiation does not happen in one leap. A pluripotent stem cell moves through a series of developmental stages before becoming a specialized cell. Each stage introduces new decisions and new opportunities for the biology to move in an unintended direction. Trailhead therefore develops differentiation protocols through a sequence of iterative experiments.
At each stage, the platform generates a broad field of data. Scientists then use that information to identify the path most likely to lead toward the desired cell type. The next experiment narrows the field further. Over time, the team follows what Trivedi called the “data trail” through the developmental lineage.
The mathematics gives scientists access to a wider range of possible paths. Biological expertise must then determine which path to pursue.
“Scientists have access to all the directions,” Trivedi said. “It is up to them and their subject-matter expertise to figure out which specific direction they are interested in.”
This is an important distinction. HD-DoE is not an automated answer generator. It does not independently decide what biology means. Instead, it creates a more complete map and it’s up to the scientists to navigate it.
The scientists’ job becomes interpreting that map by choosing which paths to pursue. These paths are then built by transforming thousands of experimental observations into a protocol that tells the scientists what to add, when to add it, and in what order so the same high-quality cells can be produced consistently. Then validation of the pathway predictions and refinement of the results are continuously done until there are robust, reproducible protocols.
Antonchuk explained, “Once we find the conditions that give us the gene expression pattern we want, we have to confirm the cells' identity and functionality by other methods. We then confirm reproducibility and stress-test the protocols to get them to the stage where they can be handed over to manufacturing."
More Than Software
Because Trailhead uses robotics, gene expression measurements, mathematical modeling, and machine learning, it may be tempting to describe HD-DoE as software. But the Trailhead team sees the company differently.
Llewellyn described the company as “more of a platform - one that can be used to make many different cell types over and over again.”
Trivedi adds, “Even though we are perceived as a biology and life sciences company, underneath we are a mathematics, software, and data-generating company.”
But HD-DoE is more than any one of those components. The robotics allow the team to perform experiments consistently and at higher throughput while the assays convert biological responses into measurable data. The mathematics identifies relationships within that data. Finally, the scientists interpret those relationships and design the next experiment. Together, those elements form a repeatable discovery system.
Standardization was built into the technology from the beginning. If experiments are performed differently each time, the resulting data cannot be compared reliably or used to build increasingly useful models.
“You cannot scale without standardization,” Trivedi said.
That principle influenced how Trailhead performed experiments, generated data, and incorporated automation. As early as 2013 and 2014, the team was already using robotics, not only to increase throughput, but to reduce variation and build a foundation that could eventually support scale. Each experiment could then contribute to more than one immediate project. Its data became part of a growing body of knowledge that could inform future experimental design. The platform did not simply produce results. It learned from its own history.
From Research Project to Platform Company
The early technology was ambitious, but turning it into a sustainable company required another set of decisions. Initially, Trailhead operated through a service-oriented model. Because each project required researchers to commit a substantial amount of funds from their budget, they were understandably cautious about trying a new and unproven approach. The team began with several demonstration projects, sharing costs to de-risk the technology and show what the technology could do. Between 2014 and 2017, those projects helped establish proof of concept. But they also revealed something larger.
Trailhead had not simply developed a way to complete individual differentiation projects. It had built a discovery engine capable of generating new protocols repeatedly. That distinction eventually changed the company’s direction. Rather than operating primarily as a service provider, Trailhead began shifting toward a product model by using HD-DoE to build its own portfolio of differentiated human cells, media, kits, and related research tools.
When David Llewellyn was later recruited to lead the company into commercialization, it was the platform itself that convinced him to transition from his role at STEMCELL Technologies.
“It was really the underlying technology,” Llewellyn said. “The technology was very compelling.”
His confidence did not come from a pitch alone. Before joining Trailhead, he reviewed the academic publications generated around the system and saw what he considered strong evidence that the approach worked.
“They had really good proof of concept that the technology was good,” he said.
That proof shaped how Llewellyn understood the company’s potential. He said, “That technology [HD-DoE] allows us to develop many different cell types and many different types of products.”
The platform itself had always been broadly applicable. The early team considered other potential uses beyond stem cell differentiation. But a framework capable of exploring many directions must still choose where to focus. Trailhead has chosen human cell differentiation. It’s the problem for which the technology had originally been conceived and a problem large enough to spend a lifetime trying to solve.
The Story of an Idea
This is what makes Trailhead more than just an interesting concept. It is the story of an idea that attracted mathematicians, developmental biologists, engineers, commercial leaders, investors, and Ohio institutions around a single question. Jensen conceived a system that could test biology beyond the limits of ordinary hypotheses, Ohio funded the first attempt, and a multidisciplinary team built the machinery, mathematical framework, experimental processes, and accumulated knowledge needed to make it work.
Trivedi and the early team helped translate the founding scientific vision into a repeatable approach, one in which experimental biology and mathematical analysis continually challenge and refine one another. More than a decade later, the founding question remains.
Trailhead is still asking how the hidden relationships inside biological systems can be made visible. Only now, it has built a system designed to search for them. HD-DoE® was not created to produce a single cell type, rather it was created to build a new way of understanding, and ultimately directing, the biological processes that determine cell fate. Trailhead is not trying to change biology itself, instead it is trying to change how scientists understand and control it.
Zakrzewski, W., Dobrzyński, M., Szymonowicz, M. et al. Stem cells: past, present, and future. Stem Cell Res Ther 10, 68 (2019). https://doi.org/10.1186/s13287-019-1165-5
Jensen, J. Changing The Scientific Process: A Decade of High-Dimensional Design-ofExperiments (HD-DoE®) at Trailhead. Trailhead Biosystems - Trail Notes, (2024). https://trailbio.com/ceo-note-july-2024/
Credits
Photography by Nicholas Campbell
Key Takeaways
A different approach to a complex problem: Trailhead Biosystems was founded on the idea that traditional, hypothesis-driven research alone may not be enough to unravel the complexity of human cell differentiation.
The origins of HD-DoE®: Founder Jan Jensen envisioned applying high-dimensional experimental design to biology, with early support from Ohio's Third Frontier program helping transform that vision into a working research platform.
Where mathematics meets biology: Trailhead's approach combines experimental biology, robotics, mathematical modeling, and scientific expertise to examine how multiple factors interact to influence cell fate.
From experimental data to reproducible protocols: Rather than simply testing individual conditions, HD-DoE helps scientists identify promising combinations of factors, refine differentiation processes, and develop consistent methods for producing specific human cell types.
A platform built for possibility: What began as an ambitious research concept has evolved into a technology platform designed to accelerate the development of human cell models for research, drug discovery, and other biomedical applications.
Learn more:
About Trailhead Biosystems
Trailhead Biosystems is a Cleveland-based biotechnology company developing human cell models for drug discovery, disease research, and regenerative medicine. Its proprietary HD-DoE® platform combines high-dimensional experimental design, computational modeling, and biological validation to improve the reproducibility and control of stem cell differentiation. Learn more about Trailhead here: https://trailbio.com/.
About The Biotech Beacon
The Biotech Beacon is an independent publication dedicated to illuminating biotechnology, life sciences, and healthcare innovation across Ohio and the Midwest. Through in-depth reporting, founder interviews, and ecosystem analysis, it connects scientists, entrepreneurs, investors, and curious readers to the people and technologies shaping the future of healthcare.
Share Your Story
Every biotechnology company has a story worth telling. If your organization is developing innovative science, building new technologies, or advancing healthcare across Ohio and the Midwest, I'd love to learn more. Whether you're a startup, an established company, an incubator, or a research organization, if you're building something that deserves to be seen, from groundbreaking science to the people making it happen, I'd love to hear from you. Reach out at smeade@thebiotechbeacon.com or connect with me on LinkedIn.
Comments