From proof of concept to platform: How Dr. Zilu Ye is scaling single-cell proteomics for the real world
“If you truly enjoy research and see it as a long-term pursuit, then it is much more like a marathon than a sprint.”
Dr. Zilu Ye, Principal Investigator at the Institute of Systems Medicine at the Chinese Academy of Medical Sciences, has spent his career developing mass spectrometry–based proteomics tools that don’t just push technical boundaries, but translate into real biological and clinical insight. His lab’s work is, as he puts it, both technology-driven and question-oriented: advancing the sensitivity and throughput of proteomics on one hand, and ensuring those advances move beyond model systems into the complexity of real tissues and disease on the other. Now, with a new high-throughput single-cell proteomics platform capable of profiling more than 10,000 cells per day, and an eye on the emerging concept of the AI-driven virtual cell, Zilu is working to establish proteins – the “direct executors of cellular function” – as an indispensable layer in how we understand biology.
A number on a screen
Zilu’s trajectory from methodological curiosity to potential world-first results didn’t follow a straight line. There’s a rare moment in science when a number appears on a screen and something shifts in what you believe is possible.
For Zilu that moment came when a Spectronaut analysis finished running and the protein count from a single cell came back above 5,000. “Before seeing the data, my own expectation was around 4,000 proteins, which I already considered an excellent result,” he says. “But when we saw that the number exceeded 5,000, it was mind-blowing.”
What made it more memorable was the reaction from collaborators. When he shared the results, the initial response was almost always disbelief.
“That transition – from not daring to expect such an outcome to seeing results that exceeded everyone’s expectations – is a rare and rewarding experience in research.”
The field was highly competitive, with multiple groups continuously pushing the limits of proteome depth, and Zilu’s lab was, quite possibly, the first to break the 5,000-protein barrier from a single cell. Not long before, Zilu had not believed that single-cell proteomics would become a truly viable approach in the near term, let alone broadly applicable. The limitations were obvious: vanishingly small sample amounts, weak signal, unpredictable system stability. The goal of characterising the full protein complement of a single cell had the quality of a thought experiment – theoretically interesting, practically remote.
What changed his mind was not a single breakthrough but a convergence: better sample preparation, higher-performance chromatographic columns, and a new generation of mass spectrometry platforms. The cumulative effect, he says, amounted to “a system-level leap” – a point at which individually incremental advances combined into something qualitatively different. The 5,000-protein result was not just a number. It was confirmation that single-cell proteomics had crossed from proof-of-concept into a platform that could actually be used and scaled.
Finding the field
Zilu’s path into proteomics began, as many research careers do, with a choice made under some uncertainty. In his final undergraduate year, around 2010–2011, he was preparing for graduate training and facing the need to select a specific direction – a shift, he notes, from the generalism of undergraduate education to something much more deliberate. He joined the Institute of Biophysics at the Chinese Academy of Sciences, and found himself drawn to mass spectrometry: precise, technically demanding, and connected in an unusually direct way to biological questions.
What struck him in particular was that mass spectrometry–based proteomics offered something rare: a genuinely global view of protein expression and regulation across the cell.
“This initial fascination, driven largely by the technology itself, shaped my research trajectory and has continued to influence my work up to the present,” he says.
What has deepened that fascination over time is the realisation that proteins are not just measurable entities but the direct executors of cellular function – the layer at which biology actually happens.
The first paper: a comprehensive learning process
Graduate training brought both opportunity and difficulty. “At the time, it was honestly quite stressful,” Zilu recalls. Beyond the technical challenges of the project itself – applying data-independent acquisition (DIA) to glycoproteomics, a field defined by molecular complexity – he was simultaneously completing his PhD thesis, preparing for his defence, and thinking about the next step in his career. “Looking back, I think this is a fairly typical situation for many PhD students,” he says, “but at the time it certainly felt overwhelming.”
The combination of problems was formidable. DIA was still in its early stages of development, and glycoproteomics placed additional demands on data interpretation and method optimisation, meaning that experimental design, data acquisition, and downstream analysis all presented significant obstacles at once. There were moments where Zilu felt in over his head, “especially when experiments didn’t work as expected or when I was struggling to interpret complex data.” On top of that, he was writing a scientific paper for the first time, without the experience to know how to structure an argument, present data clearly, or respond to reviewers.
What got him through was a combination of close guidance from his supervisors and a willingness to keep iterating, even when progress felt slow.
“I wouldn’t say I handled everything perfectly,” he says, “but I do think I had a relatively resilient mindset. I tried to stay optimistic and focus on moving forward step by step.”
The paper was eventually published in a well-regarded journal, but the equally durable outcome was the confidence it built. “Seeing everything come together – not just in terms of the results, but also in gaining confidence in how to approach research – was incredibly rewarding.”
Building for scale
That methodological rigour has carried through into the work his lab is producing now. Their most recent platform, centred on a system called SPRINT, represents an attempt to solve one of the persistent bottlenecks in single-cell proteomics: scale.
SPRINT is an AI-powered single-cell printing platform capable of stably preparing more than 10,000 single cells per day, an order-of-magnitude increase over existing systems. Crucially, this gain in throughput does not require trading away sensitivity; the platform still identifies over 6,000 proteins at the single-cell level. To complement it on the analytical side, Zilu’s lab designed a dual-spray tandem direct injection (TDI) LC–MS architecture that parallelises the non-analytical steps in liquid chromatography, effectively doubling mass spectrometer utilisation and enabling approximately 168 single-cell analyses per day without reducing proteome depth.
The architecture of this system was designed from the outset around specific chromatographic columns – in this case, IonOpticks Aurora Rapid 5 cm columns, whose stability, proteome coverage, and performance under high-throughput, low-input conditions are central to how the system functions. In work pushing sensitivity to its limits, Zilu notes, differences between columns become significantly amplified.
“The chromatographic system is equally as critical as the mass spectrometry platform or sample preparation, yet sometimes underappreciated,” he says. “For these types of cutting-edge applications, the choice of column is no longer simply an optimisation parameter – it has effectively become an integral part of the overall system design.”
Beyond cultured cell lines, the lab has also established standardised workflows for tissue-derived cells and constructed a cross-organ single-cell proteomic atlas – demonstrating, as Zilu puts it, “the applicability of the platform in more complex biological contexts.” The significance of that step derives from the reality that moving from controlled, homogeneous cell lines to the messiness of real tissue is where many platforms falter. For Zilu, it represents something more than a technical milestone: a step toward transitioning single-cell proteomics from proof-of-concept studies with limited sample sizes into a platform capable of supporting systematic, large-scale biological investigations.
Progress in a field that moves quickly requires knowing what to compare yourself against. This is something Zilu has thought about carefully. “I think I actually came to this realisation relatively early, probably during my master’s studies,” he says. “But the feeling became much stronger over time, shaped by my own experiences, including both the highs and the lows, as well as by observing many of my peers going through similar paths.” Earlier in his career, especially during his PhD and early postdoc years, he was still quite influenced by competition and peer pressure. In a fast-moving field like proteomics, he notes, it is very easy to associate progress with speed: how quickly you publish, how early you report a result. “Over time, I realised that this mindset is difficult to sustain in the long run.”
“If you truly enjoy research and see it as a long-term pursuit,” he says, “then it is much more like a marathon than a sprint.”
He describes his approach now as one of continuously improving a personal best, rather than positioning against others – a framing that treats research as an internal standard rather than a competitive ranking. The “peer pressure” he mentions is real, especially early in a career, when publication timing and intermediate results can feel like proxies for worth. But short-term outcomes, he argues, are often shaped more by environment and opportunity than by ability; publication timing and intermediate results “do not fully define one’s long-term trajectory.” What matters more is developing your own pace and your own standards.
His advice to his younger self would be to “try to stay calm and develop a resilient mindset. In the long term, that stability is far more valuable than short-term gains.”
What accelerates progress, he argues, is not optimising any single component but recognising when multiple advances converge into something genuinely new. He describes this as a “systems-level” shift. A striking example from his own experience was testing a prototype of the Orbitrap Astral mass spectrometer, before it was officially named or released. When he observed it acquiring more than 200 MS2 spectra per second in a stable manner – a scanning rate that would have been almost unimaginable a few years prior – the implication was immediate. “It made it very clear,” he says, “how advances in instrumentation can fundamentally redefine what is possible in the field.”
Towards the virtual cell
That same systems-level thinking animates what Zilu sees as the most ambitious direction in proteomics right now and, if resources allowed, what he would pursue most aggressively.
The broader field, he notes, has already made significant strides – not just in sensitivity and throughput, but in addressing increasingly complex biological questions: post-translational modifications, protein–protein interactions, and the integration of proteomics with other omics layers such as transcriptomics and metabolomics. “This convergence,” he says, “allows us to generate richer datasets while enabling a more systems-level understanding of biological processes.” In other words, moving beyond individual molecules or pathways to understand how different molecular layers interact to define cellular states and behaviours as a whole. The overarching trend, in his view, is proteomics evolving from a purely analytical technology into a comprehensive platform for systems biology.
At the centre of that evolution, for Zilu, is the concept is the AI-driven Virtual Cell: a computational model capable of systematically predicting and simulating cellular states and behaviours. The virtual cell, he explains, can be thought of as a predictive model of a cell – one that could simulate how a cell responds to perturbations such as disease or drug treatment. The ambition is significant and increasingly concrete: a mature virtual cell could help predict how a tumour might evolve or respond to a specific intervention, or identify which patients are likely to respond to a given therapy – the kind of insight that sits at the heart of precision medicine – without requiring a physical experiment at every step.
At present, most virtual cell efforts lean heavily on transcriptomic data – the layer that captures which genes are being expressed. The proteomic layer, which captures what the cell is actually doing rather than what it is instructed to do, remains significantly underrepresented. Part of the reason is practical: compared to transcriptomics, proteomics still faces greater challenges in throughput, robustness, and standardisation. But Zilu sees those as surmountable and, once overcome, an opening.
“Proteins are the direct determinants of cellular function,” he says, “yet they remain significantly underrepresented in current virtual cell models.”
He sees this not only as a gap but as an opportunity – one that single-cell proteomics, as it continues to mature, is increasingly well-positioned to fill.
The broader trajectory he describes for proteomics is one of transformation: from a measurement technology into a platform for systems biology, and from basic research into the kinds of clinical applications that affect patients directly: disease stratification, target discovery, precision medicine. “We are at a particularly exciting and pivotal moment,” he says, “where accumulated technological advances are beginning to translate into capabilities that can genuinely reshape how we study biology.”
Perhaps the clearest expression of his outlook is this: “In proteomics, some goals we once assumed would take a long time to achieve may, in fact, arrive much sooner than expected.”
For a researcher who once doubted single-cell proteomics would become viable in the near term, before helping to make it so, that’s a hard-won belief.
Detailed FAQs
I am currently a principal investigator at the Institute of Systems Medicine, Chinese Academy of Medical Sciences. My research focuses on mass spectrometry–based proteomics, with a particular emphasis on single-cell proteomics and methodological development.
Broadly speaking, our work is both technology-driven and question-oriented. On the one hand, we aim to continuously push the boundaries of sensitivity and throughput in proteomics. On the other hand, we are equally interested in how these technologies can be translated into real biological and clinical contexts—especially in complex systems such as tissues and diseases—rather than remaining confined to model systems.
My initial interest in this field was largely driven by the technology itself. Mass spectrometry is not only a highly precise analytical tool, but also one that connects directly to biological questions. As my research progressed, I came to appreciate that proteins, as the direct executors of cellular function, provide a uniquely information-rich and biologically relevant layer of insight. This realization is ultimately what has kept me deeply engaged in this field.me
My interest in proteomics began during my final year of undergraduate studies, around 2010–2011, when I was preparing to pursue graduate training. Compared to the relatively general education at the undergraduate level, graduate school requires a much more deliberate choice of a specific research direction, which was a pivotal moment for me.
At that time, I joined the Institute of Biophysics at the Chinese Academy of Sciences, a highly regarded research institute in life sciences. As I explored different research areas, I found myself naturally drawn to mass spectrometry. On the one hand, it is an exceptionally precise analytical tool with strong technical appeal; on the other hand, mass spectrometry–based proteomics enables a global view of protein expression and regulation, which I found particularly striking.
In many ways, this initial fascination—driven largely by the technology itself—shaped my research trajectory and has continued to influence my work up to the present.
ulThis was less a single moment and more a process that led to a clear shift in my perspective, and in many ways aligns with views expressed by Professor Matthias Mann. A few years ago, I did not believe that single-cell proteomics would become a truly viable—let alone broadly applicable—approach in the near term. From a technical standpoint, there were obvious limitations, including extremely low sample amounts, weak signal intensity, and challenges in system stability.
However, over the past few years, the pace of progress in this field has far exceeded my expectations. With continuous improvements in sample preparation strategies, advances in chromatographic performance—such as the development of higher-performance columns—and the emergence of next-generation mass spectrometry platforms, the entire technological landscape has undergone what I would describe as a “system-level leap.”
For me, a particularly striking, even mind-blowing moment was when we were able to consistently identify more than 5,000 proteins from a single cell. This was not just a numerical milestone; more importantly, it marked the transition of single-cell proteomics from a conceptual or proof-of-principle stage into a genuinely usable and scalable platform.
It also led me to rethink a broader point: in proteomics, some goals we once assumed would take a long time to achieve may, in fact, arrive much sooner than expected, driven by the convergence of multiple technological advances.
lIf I had to choose one particularly memorable experience, it would be our breakthrough in single-cell proteomics. At that time, we were likely the first group to identify more than 5,000 proteins from a single cell.
The field was highly competitive, with multiple groups continuously pushing the limits of proteome depth. To be honest, before seeing the data, my own expectation was around 4,000 proteins, which I already considered an excellent result. But when the Spectronaut analysis was completed and we saw that the number exceeded 5,000, it was genuinely striking and extremely exciting.
What made it even more memorable was the reaction from our collaborators. When we shared the results, the initial response was almost always disbelief. That transition—from not daring to expect such an outcome to seeing results that exceeded everyone’s expectations—is a rare and rewarding experience in research.
Another moment that left a strong impression on me was the first time I tested the Orbitrap Astral (before it was officially named). When I directly observed that it could acquire more than 200 MS2 spectra per second in a stable manner, the impact was quite profound. It made it very clear how advances in instrumentation can fundamentally redefine what is possible in the field.
We recently developed an end-to-end platform for single-cell proteomics, centered around the SPRINT system and a complementary dual-spray TDI LC–MS architecture.
SPRINT is essentially an AI-powered single-cell printing platform that enables the stable preparation of more than 10,000 single cells per day—representing an order-of-magnitude increase in throughput compared to existing systems. Importantly, this gain in throughput does not come at the expense of sensitivity; we are still able to identify over 6,000 proteins at the single-cell level.
On the analytical side, we designed a dual-spray TDI system that parallelizes the non-analytical steps in LC, effectively doubling MS utilization efficiency. This allows us to reach a throughput of approximately 168 single-cell analyses per day without compromising proteome depth.
Beyond these technical advances, a key aspect of this work is that we moved beyond cultured cell lines. We established standardized workflows for tissue-derived cells and constructed a cross-organ single-cell proteomic atlas, demonstrating the applicability of the platform in more complex biological contexts.
From my perspective, the significance of this work lies not in any single performance metric, but in addressing a long-standing bottleneck in the field—scalability. In other words, it represents a step toward transitioning single-cell proteomics from proof-of-concept studies with limited sample sizes to a platform capable of supporting systematic, large-scale biological investigations.
One of the most challenging experiences I encountered was during my PhD, when I worked on my first research paper.
At the time, it was honestly quite stressful. Beyond the technical challenges of the project itself, I was also trying to complete my PhD thesis, prepare for my defence, and think about the next step in my career. Looking back, I think this is a fairly typical situation for many PhD students, but at the time it certainly felt overwhelming.
We were attempting to apply data-independent acquisition — DIA, which was still relatively new in proteomics — to glycoproteomics, a field that is inherently more complex and places additional demands on data interpretation and method optimisation. As a result, we faced considerable challenges across experimental design, data acquisition, and downstream analysis, all at once. There were definitely moments when I felt somewhat in over my head, especially when experiments didn’t work as expected or when I was struggling to interpret complex data.
At the same time, this was my first paper, so I also lacked experience in scientific writing — from structuring the data and clearly articulating the scientific questions, to responding to reviewers’ comments.
What helped me get through that period was a combination of close guidance from my supervisors and a willingness to keep iterating, even when progress felt slow. I wouldn’t say I handled everything perfectly, but I do think I had a relatively resilient mindset — I tried to stay optimistic and focus on moving forward step by step.
In the end, we successfully completed the work and published it in a well-regarded journal. But more than the result itself, seeing everything come together — including gaining confidence in how to approach research — was incredibly rewarding. I believe this kind of experience is something many early-career researchers can relate to, and in hindsight, it was an important part of my development.
Looking back, one thing I wish I had realised earlier is that if you truly enjoy research and see it as a long-term pursuit, then it is much more like a marathon than a sprint.
When did that realisation form for you?
I think I actually came to this realisation relatively early — probably during my master’s studies. But the feeling became much stronger over time, shaped by my own experiences, including both the highs and the lows, as well as by observing many of my peers going through similar paths. Earlier in my career, especially during my PhD and early postdoc years, I was still quite influenced by competition and peer pressure, particularly in a fast-moving field like proteomics. It is very easy to associate progress with speed — how quickly you publish, or how early you report a result. Over time, I realised that this mindset is difficult to sustain in the long run.
Short-term outcomes are often shaped more by environment and opportunity than by ability — the timing of publications and intermediate results do not fully define one’s long-term trajectory. What matters more is developing your own pace and your own standards. I tend to think of research as a process of continuously improving one’s personal best, rather than competing directly with others.
If there is one practical takeaway, it would be to try to stay calm and develop a resilient mindset. In the long term, that stability — that sense of rhythm and consistency — is far more valuable than short-term gains.
What do you consider the most exciting recent development in proteomics, and how might it impact future research or applications?
Over the past few years, proteomics has advanced remarkably, and importantly, across multiple dimensions. From a technical perspective, both throughput and sensitivity have improved significantly, enabling high-quality measurements from increasingly small sample amounts, including at the single-cell level. At the same time, on the application side, there has been growing progress in addressing more complex biological questions, such as post-translational modifications (PTMs) and protein–protein interactions (PPIs).
In parallel, proteomics is becoming increasingly integrated with other omics layers—such as transcriptomics and metabolomics—as well as with emerging AI-based approaches. This convergence not only allows us to generate richer datasets, but also enables a more systems-level understanding of biological processes. By “systems-level,” I mean moving beyond individual molecules or pathways to understand how different molecular layers interact to define cellular states and behaviours as a whole.
In my view, the overarching trend behind these developments is that proteomics is evolving from a purely analytical or measurement technology into a comprehensive platform for systems biology. This shift will likely continue to enhance its role in fundamental research, while also expanding its impact in translational and clinical settings. Concrete examples include predicting how a tumour might evolve or respond to a specific intervention, or identifying which patients are likely to respond to a given therapy — the kind of insight that sits at the heart of precision medicine.
Overall, I believe we are at a particularly exciting and pivotal moment for the field—where accumulated technological advances are beginning to translate into capabilities that can genuinely reshape how we study biology.
I was introduced to IonOpticks columns relatively early on, during my time in Professor Jesper Olsen’s laboratory, so I would consider myself among the early users.
From a practical standpoint, they have consistently impressed me in terms of both stability and chromatographic performance. This becomes particularly important in applications such as high-throughput workflows, short-gradient separations, and single-cell proteomics, where chromatographic performance can have a substantial impact on overall data quality. To be quite candid, many of our best results have been obtained using IonOpticks columns. When people ask me about the key factors for achieving high-quality single-cell proteomics, I often emphasise that it is not only about the mass spectrometry platform or sample preparation — the chromatographic system is equally critical, yet sometimes underappreciated.
A concrete example is our recently developed dual-column LC system, which was designed from the outset around the specifications of IonOpticks 5 cm columns, including the optimisation of gradient conditions, flow rates, and overall workflow integration. In such extremely low-input scenarios, where system stability and sensitivity are pushed to their limits, differences between columns become significantly amplified — and in our experience, IonOpticks columns consistently deliver more stable performance and higher proteome coverage than alternatives we have worked with.
For these types of cutting-edge applications, the choice of column is no longer simply an optimisation parameter — it has effectively become an integral part of the overall system design.
If funding weren’t an issue, what’s the most ambitious experiment you’d run?
If funding were not a constraint, the direction I would most like to push forward is the integration of single-cell proteomics into the framework of the AI-driven Virtual Cell (AIVC).
I see the virtual cell as one of the most exciting and potentially transformative directions in biomedicine today — a predictive model of a cell capable of simulating how it responds to perturbations such as disease or drug treatment. Its goal is to systematically predict and simulate cellular states and behaviours. Concrete applications could include predicting how a tumour might evolve or respond to a specific intervention, or identifying which patients are likely to respond to a given therapy — the kind of insight that sits at the heart of precision medicine.
However, most current efforts rely heavily on transcriptomic data, while the proteomic layer — arguably the most direct determinant of cellular function — remains significantly underrepresented. Compared to transcriptomics, proteomics still faces greater challenges in throughput, robustness, and standardisation. But once these bottlenecks are overcome, proteomics could provide a critical missing dimension for the virtual cell.
With sufficient resources, I would focus on two fronts: further developing high-throughput, scalable single-cell proteomics technologies, and systematically integrating these data into virtual cell models to establish a protein-centric framework for describing cellular states.
In the long term, I believe this will not only complement existing multi-omics approaches, but could play a key role in moving the concept of the virtual cell toward practical implementa
To provide the best experiences, we use technologies like cookies to store and/or access device information. Consenting to these technologies will allow us to process data such as browsing behaviour or unique IDs on this site. Not consenting or withdrawing consent, may adversely affect certain features and functions.
Functional
Always active
The technical storage or access is strictly necessary for the legitimate purpose of enabling the use of a specific service explicitly requested by the subscriber or user, or for the sole purpose of carrying out the transmission of a communication over an electronic communications network.
Preferences
The technical storage or access is necessary for the legitimate purpose of storing preferences that are not requested by the subscriber or user.
Statistics
The technical storage or access that is used exclusively for statistical purposes.The technical storage or access that is used exclusively for anonymous statistical purposes. Without a subpoena, voluntary compliance on the part of your Internet Service Provider, or additional records from a third party, information stored or retrieved for this purpose alone cannot usually be used to identify you.
Marketing
The technical storage or access is required to create user profiles to send advertising, or to track the user on a website or across several websites for similar marketing purposes.
To provide the best experiences, we use technologies like cookies to store and/or access device information. Consenting to these technologies will allow us to process data such as browsing behaviour or unique IDs on this site. Not consenting or withdrawing consent, may adversely affect certain features and functions.
Functional
Always active
The technical storage or access is strictly necessary for the legitimate purpose of enabling the use of a specific service explicitly requested by the subscriber or user, or for the sole purpose of carrying out the transmission of a communication over an electronic communications network.
Preferences
The technical storage or access is necessary for the legitimate purpose of storing preferences that are not requested by the subscriber or user.
Statistics
The technical storage or access that is used exclusively for statistical purposes.The technical storage or access that is used exclusively for anonymous statistical purposes. Without a subpoena, voluntary compliance on the part of your Internet Service Provider, or additional records from a third party, information stored or retrieved for this purpose alone cannot usually be used to identify you.
Marketing
The technical storage or access is required to create user profiles to send advertising, or to track the user on a website or across several websites for similar marketing purposes.
You can now build your quote request online 
