This January, we moved the first candidate medicine that emerged from that plan into large-scale manufacturing. We also discovered another payload that restores multiple youthful functions, built preclinical models in our third therapeutic program, and finalized an AI system that doubles our rate of discovery. These results have surpassed even our optimistic projections.

Recent highlights:

• +1 payload that restores multiple youthful functions in old hepatocytes

• 0 → 20% manufacturing progress our first large-scale batch of drug candidate

• 0 → 4 candidate pharmacodynamic biomarkers

• +50 payloads that make old hepatocytes look young

• +19 payloads that rescue regulatory activity in T cells

• +1 preclinical model of renal endothelial cell age

• 2X faster payload discovery with multi-task in silico reprogramming models

Restoring youthful metabolic function

Epigenetic marks on our DNA and surrounding proteins control which genes are turned on and off in each cell. Aging degrades these marks, and old cells lose the ability to perform obligate functions as a result. As one example, hepatocytes (Greek/biologist for “liver cell”) are less regenerative in the face of injury, less resilient to external damage, and less able to process metabolites.

NewLimit is discovering reprogramming payloads that rewrite these marks back to their youthful state. Each of our payloads activates a set of transcription factor (TF) genes that remodel the epigenome. In principle, payloads that effectively reprogram cell age could restore multiple youthful functions in old cells.

In February, we discovered another payload that displays this form of pleiotropy (Greek/biologist for “does multiple things”) in hepatocytes. After reprogramming with these TFs, old hepatocytes are both more resilient to damage from alcohol and more regenerative in the face of injury. The payload appears to have similar effects in both mouse and human hepatocytes, suggesting the biology is conserved across species.

When we began our discovery campaigns years ago, it was unclear if this form of multi-functional recovery was possible. We’ve now found multiple payloads that achieve this effect. Cell age can be reprogrammed while preserving cell type, and only a small set of transcription factors are required.

Inspired by results like these, we’ve been scaling our humanized liver reprogramming screens to discover new payloads. In the past months, we’ve landed 4 large-scale screens that revealed dozens of new payloads that restore youthful features in old human hepatocytes. These hits not only restored youthful gene expression, but also rescued regenerative function in this in vivo setting.

Therapeutics development typically involves discovering many promising candidates at the early stage, then carrying a small number of candidate medicines (e.g. 1-10) forward to clinical trials. In this setting, the quality of the top hits matters as much if not more than the number of hits we discover in total.

We’ve found that as we scale our discovery screens, the quality of the top k hits likewise continues to improve. This biological scaling law resembles the scaling properties of artificial intelligence systems and follows a log-linear trend. Due to this phenomenon, scaling our early stage discovery efforts is likely to result in better medicines when we enter development.

Artificial intelligence systems accelerate reprogramming discoveries

Discovering TF payloads that reprogram cell age is challenging because the number of possible payloads is uncountably large. There are about 10¹⁶ TF sets we might test, many orders of magnitude more than we can tractably test in the laboratory. In addition to building industrial scale experimental systems, we need to prioritize the experiments we run intelligently.

Humans struggle to reason through the complex prior information required for this prioritization task at scale. By contrast, modern artificial intelligence systems can achieve superhuman performance and help prioritize the most valuable experiments given only a sparse training set.

We’ve built a system we call Ambrosia to perform these reprogramming experiments in silico. In January, we upgraded our system to operate in a multi-task fashion, predicting several effects of reprogramming on cell state and function simultaneously. These in silico experiments now allow us to prioritize payloads based on multiple axes simultaneously. For instance, we prioritize payloads that are predicted to reverse cell age while also preserving cell type.

Our multi-task Ambrosia system is highly performant across a range of cell state and function prediction tasks. When we integrate the system with a lab-in-the-loop to perform discovery campaigns, we achieve our strongest ever performance and increase our rate of discovery >2X relative to the baseline. Every time we launch a screen at NewLimit – multiple times per month – we’re able to make more discoveries per dollar as a result.

We’re working now to expand the palette of functional effects our system can both predict and design toward. Our ambition is to achieve a large increase in the translation rate of our primary screening experiments to downstream preclinical validation over the coming year.

Measuring endothelial cell age in the kidney

Our Vascular program is developing medicines to restore function in aged endothelial cells. Endothelial cells line the walls of our blood vessels, providing a transport system for nutrients and signals to flow into tissues. Endothelial cells also dictate which cells and signals can exit the circulation and enter a tissue at each juncture.

As we age, endothelial cells lose both their regenerative capacity and barrier function. The kidney is one of many tissues that suffers as a result. Kidney function declines in almost everyone with age, and age-related kidney diseases represent a massive burden on patients and the healthcare system. Kidney care accounts for roughly 0.5% of US GDP, higher than the GDP share of the Apollo Program. Our initial medicines are targeted at reprogramming renal endothelial cells to restore kidney function and address these needs.

Before we can evaluate potential reprogramming medicines, we need to develop assays to measure age-related impairments in renal endothelial cells. The ideal assay demonstrates that old cells are less functional than young cells, failing to preserve health in the face of injury or disease.

Last month, we developed one such preclinical model that induces kidney disease specifically by injuring the renal endothelium. We discovered that old animals are significantly more susceptible to mortality from this injury than young mice. Given that these injuries are specific to the endothelium, these results suggest that old renal endothelial cells are significantly more vulnerable to injury than young cells.

Paired with our novel RNA delivery chemistry, we now have the tools required to begin testing reprogramming payloads for their effect on endothelial cell age.

Elsewhere

We’ve been privileged to share the NewLimit story in several venues recently. Find us elsewhere below!

• Kleiner Perkins Builders podcast: we discuss epigenetic reprogramming, science-driven companies, and Chick-fil-A

• Biotech TV: Brad Loncar interviews us inside the lab

• Inc Magazine: NewLimit and our science are featured

Company growth

We continue to expand our team as we move toward the clinic. We’re recruiting for roles across the spectrum of seniority and across the research & development spectrum. Please reach out if you or a talented friend might be a good fit!

https://newlimit.com/careers

The science moved faster than we anticipated. We began developing our first medicines ahead of schedule in 2025. Our key accomplishments:

1. Built our 1st preclinical candidate that restores multiple youthful functions in old hepatocytes

2. Developed frontier AI systems that increase our discoveries/$ by >2X

3. Discovered 30+ reprogramming payloads that restore youthful function in old cells

4. Launched our 3rd therapeutic program focused on Vascular biology & built multiple functional assays for cell age

Discovery cadence & metrics

NewLimit is built around a set of core technologies we call our Discovery Engine. The goal of our Engine is to discover transcription factor (TF) payloads that restore youthful function in old cells by reprogramming the epigenome, allowing us to treat disease and preserve health in old patients.

Our Engine first predicts which TF payloads will be effective with AI systems, then tests thousands of payloads at a time for their effect on old human cells. From these data, we can discover payloads that make old cells look young based on their active genes and act young based on their functional performance. We then build prototype medicines from the payloads we discover and test to see if we can treat or cure a disease in animal models of aging.

In 2025, we increased our Engine throughput many fold and decreased the cost of each new discovery. The numbers themselves communicate this most clearly.

Our team managed to increase both the scale and fidelity of our Engine this year. To our knowledge, we’ve now tested >1000X as many reprogramming payloads as the rest of the field combined. We tested more payloads in 2025 than in all of our cumulative experiments from 2022-2024. At the same time, we expanded our repertoire of gold-standard preclinical models and improved the fidelity of our scaled screens to discover functional hits, dramatically reducing the cost per discovery.

As a result of these improvements, we have now discovered 16 payloads that treat disease in gold-standard animal models, 36 payloads that restore youthful function in cells, 600+ that make old cells look young based on gene expression, and 0 → 1 preclinical candidate.

Entering early development

Therapeutic discovery transitions into development once a single molecular entity known as a preclinical candidate is chosen for in-depth evaluation. We imagined at the founding of NewLimit that it would take 5+ years to create one, but this year we delivered a candidate after only 3 years of operations.

Our bar to declare a preclinical candidate is higher than in many traditional settings. We will only advance prototype medicines to candidates if they are both safe and restore multiple youthful functions, suggesting that cell age is genuinely reversed by reprogramming . By contrast, many traditional therapeutic campaigns focus on a single target biology to rescue. Cell age is associated with health & disease progression in humans, so we hope that our more stringent criteria improve our chances of success in the clinic.

Our first candidate is based on a payload with pleiotropic activity, making old hepatocytes both look young and act young across multiple dimensions. We first discovered the payload in a humanized liver screen where it restored a youthful gene expression profile in old human hepatocytes and improved regenerative function. We then tested if this payload could rescue more than just this one function in this one context.

Old hepatocytes become more sensitive to damage and lose their ability to regenerate liver tissue after damage has occurred in animals. We found that our candidate payload not only restores resilience and prevents damage from alcohol, but also restores youthful regenerative capacity in a surgical injury model. The payload was safe in animals even at doses far exceeding the therapeutic level. Neither liver toxicity or neoplasia (e.g. tumors) were observed.

These rejuvenating effects convinced us that this payload warranted progression into development.

Our initial prototype LNP-RNA compositions for the candidate payload weren’t meaningfully optimized. We therefore executed a lead optimization campaign to improve both the potency and specificity of the payload prior to later stage studies. Our team established an RNA engineering process that yielded both a 1.6X increase in payload potency and an 8X increase in specificity for hepatocytes. These improvements are likely to yield a safer, more effective medicine.

Creation of a preclinical candidate represents a meaningful milestone for NewLimit. More broadly, we believe it’s a milestone for the partial reprogramming field as one of the first therapeutic programs to reach this stage.

AI systems accelerate therapeutic discovery

Our reprogramming payloads are composed of combinations of transcription factors. There are >10¹⁶ possible combinations we might reasonably test, thousands of times more than the number of stars in the Milky Way galaxy. No matter how performant our experiments are in the world of atoms, we will never be able to exhaustively search this space. In order to discover an optimal reprogramming medicine, we need to effectively prioritize which hypotheses we test before ever entering the laboratory.

Only a few years ago, this problem was intractable. Human intuition is insufficient to reason through a hypothesis space of this scale. Modern artificial intelligence systems by contrast can incorporate a plurality of prior knowledge to design experiments that maximize our rate of discovery.

This year we introduced Ambrosia, the first AI system that can accelerate the design of reprogramming payloads. Ambrosia builds atop knowledge captured in foundational models of molecular biology and natural language, initializing our system with a strong prior derived from both human languages and nature’s languages. Our system first learns to predict the effect of arbitrary reprogramming payloads on both the state and function of old cells.

Given this discriminative model, Ambrosia can design reprogramming payloads that induce a target cell state or phenotype. The design process can learn continually from sequential experiments, allowing the system to iteratively navigate through the hypothesis space toward an optimal payload.

Performance on the initial prediction task is a special case of perturbation prediction (“virtual cell”) modeling, and Ambrosia achieved state-of-the-art performance relative to top baselines in the field. On the latter design task, our latest Ambrosia models improve our rate of discovery by >2X relative to baselines, directly increasing the number of discoveries we make per dollar invested. We believe this is among the first AI systems that has been integrated effectively into a large scale target discovery process.

The complexity of human pathology far exceeds the sophistication of most therapeutics. Our medicines have been constrained by our intelligence. It’s been intractable to design therapies that match biology’s complexity using our traditional systems of discovery. Reprogramming medicines are only now entering the realm of the possible due to the advent of AI systems that remove this intelligence constraint. If successful, the design of reprogramming medicines that extend human health will represent one of the more dramatic impacts of AI on our world.

Diversifying lineages

NewLimit’s ultimate ambition is to create reprogramming medicines to restore function across diverse cell types in the body. Our progress in hepatocytes and T cells this year convinced us to diversify our therapeutic bets to a third cell type.

In August, we launched a Vascular program focused on restoring function in endothelial cells, the cell type that lines blood vessels. Endothelial cells are present in almost all tissues, forming the transportation network that connects organs. Perhaps unsurprisingly, endothelial aging has detrimental effects across the body in the kidney, heart, lung, and potentially even the brain. By restoring function in the endothelium, we hope to provide near-term benefits in kidney disease and long-term benefits in many tissues.

Each of our therapeutic programs requires technology to (1) test many reprogramming payloads for their effect on cell age, (2) deliver reprogramming payloads as RNA to target cell types in vivo, and (3) measure the functional impact of reprogramming on cell age.

Our team demonstrated that our existing Engine technology transferred to endothelial cells with 0 modifications in just 3 months, 2X as fast as our prior record for onboarding a program. For delivery, we formulated a new lipid nanoparticle chemistry and achieved >60% delivery to kidney endothelial cells. These delivery tools unlock in vivo experiments, and our team also developed both in vitro and in vivo models of endothelial cell age in just four months.

At the start of the year, our Vascular program was merely an idea. Today, we have assays to measure endothelial cell age, the screening technology to discover rejuvenating payloads, and the delivery tools to validate functional effects in vivo.

Molecular innovations

The scale of our discovery work in 2025 was enabled by a series of molecular inventions.

• In vivo lymphocyte delivery: Measuring the effects of reprogramming in preclinical models requires an in vivo delivery chemistry to carry our prototype medicines to the right cell types. We implemented our first targeted LNP chemistry this year and achieved highly efficient delivery to T cells, approaching state-of-the-art performance. These tools have allowed us to test T cell reprogramming payloads in vivo for the first time.

• Pooled molecular construction: Traditionally, our team relied upon a small number of arrayed molecular reactions to build reagents for our Discovery Engine screens. These steps require each hypothesis we test to exist for a time as a set of nucleic acids in an individual test tube. This year, our team developed a long-read sequencing method to eliminate these last few arrayed stages in our process, cutting the cost and turnaround time to build reagents for our screens by >5X.

• Humanized liver engineering: NewLimit invented a method to run large-scale hepatocyte screens in vivo using a humanized liver system. Over the course of 2025, we made several improvements to this system and achieved a >10X increase in the throughput/$.

No one single molecular advance underlies our Discovery Engine technology. Rather, a series of these technical improvements have compounded over the course of three years to yield a reproducible system for discovering and validating new reprogramming medicines.

2026

NewLimit’s first two years were focused on a set of fundamental research and technology problems. We built a system to search for reprogramming medicines, discovered the first payloads that can reprogram cell age, and demonstrated that restoring youthful function in old cells was possible. In 2025, we created the first medicine we hope to bring to human trials and showed that it can rescue multiple youthful functions in old cells.

2026 will be the year that we transition from purely a research enterprise to an integrated research & development organization. Effective execution in this next stage will allow us to bring reprogramming technology into human patients. If these medicines are successful, we believe they are among the most valuable products possible.

Join NewLimit

Our progress to date is the product of a talented team of scientists, engineers, & operators. We’re recruiting across the company as we enter our next stage of growth. Please reach out if you’re interested in joining us.

A few numerical highlights:

• 8X the specificity & 1.6X the potency of our lead prototype medicine

• +2 TF sets with therapeutic benefit in liver disease models

• 0 → 1 functional assays for endothelial cell age

• >20 TF sets that restore youthful phenotypes in old hepatocytes

• >3000 TF sets tested across programs

• 5X decrease in molecular construction costs

People

These results reflect the contributions of a focused team. We’ve been excited to welcome several talented scientists recently.

• Angela Boroughs joined as a Director to lead our Immunology therapeutic program. Angela previously led synthetic biology & immunology teams at ArsenalBio and directed the development of a clinical stage CAR-T cell therapy. She holds a Ph.D. in Immunology from Harvard University.

• Daniel Baca joined as a Research Associate on our Single Cell Technology team from ArsenalBio.

• Sita Chandrasekaran joined as a Scientist on our Epigenetic Editing team. She previously completed Ph.D. training with Patrick Hsu and postdoctoral work with Silvana Konnermann at the Arc Institute.

• Sean Hsieh joined as a Scientist on our Metabolism team. He previously completed a Ph.D. with Hao Zhu at the University of Texas Southwestern.

Engineering therapeutic molecules

Over the course of 2025, NewLimit has identified many reprogramming payloads that restore youthful function in old hepatocytes. In September, we selected one of these lead payloads to progress into later stages of therapeutic development.

This selection triggered the start of a lead optimization campaign. We set out to design mRNA sequences encoding our payload that (1) maximize the potency of the prototype medicine and (2) improve the selectivity of the medicine for activity in hepatocytes.

Sequences that translate from mRNA into protein more efficiently are expected to have a higher potency per unit dose. We implemented a range of sequence optimizations and evaluated their translational efficiency across a battery of different human cell types. We ultimately designed a sequence that increased the potency of our payload by 1.6X relative to our starting sequence.

mRNAs are not translated into proteins equally across cell types. Within the cytosol of the cell, a range of RNA binding proteins, RNA-guided nucleases, and the abundance of translational machinery dictate how many proteins are translated from each mRNA molecule. One cell type might make hundreds of proteins from an mRNA sequence, whereas another cell type might produce only one. We can take advantage of these differences to build a medicine that has minimal activity outside the cell type of interest, improving the safety profile.

Similar to our potency optimizations, we designed a range of sequences and eventually achieved an 8X improvement in the specificity of our prototype medicine for hepatocytes relative to off-target cell types. Combined with other optimizations, our prototype now achieves a dramatic specificity for hepatocyte activity relative to off-target cell types.

We’re now evaluating performance of the fully optimized product across a range of preclinical models and hope to progress toward formal development soon.

Restoring youthful resilience

Alongside the progression of our first assets, we continue to discover new reprogramming payloads through our Discovery Engine. In October, we validated the performance of +2 new payloads that restore youthful function, making old cells act young.

Each of these payloads originally emerged from our in vivo screens in hepatocytes. Neither has been extensively explored in prior work to our knowledge.

After discovering each payload in a screen, we formulated prototype medicines and tested their ability to protect old hepatocytes from a diet-induced injury. We then measured the amount of damage to the liver using circulating biomarkers of damage. These biomarker proteins are only produced inside hepatocytes. If we find them in the blood, it suggests hepatocytes have recently experienced damage and “leaked” these proteins into the blood. The less we detect, the healthier the liver is.

Both of these new payloads restored a youthful level of resilience to a dietary insult, dramatically lowering the amount of these damage markers we detected in the blood.

We’ve continued to discover new payloads like this month after month. These results give us confidence that our Engine can produce more and more discoveries over time.

In vivo reprogramming of T cells

Hepatic reprogramming is only one of several therapeutic applications we’re pursuing. We’ve previously discovered a range of reprogramming payloads that restore youthful cytotoxic capacity and stem-like states to old human T cells. The next stage in the development of these payloads is to test whether they can rescue function in a more complex animal model. Before we can test their effects directly, we need to build molecular tools to deliver reprogramming payloads to T cells in vivo.

Lipid nanoparticles (LNPs) are the current state-of-the-art method for delivering RNA payloads to cells in the body. They can be readily delivered to hepatocytes in the liver because hepatocytes naturally pull lipids (”fats”) from the bloodstream as part of their energy production and storage function. Delivering to other cell types requires more nuance. Typically, a targeting molecule is introduced on the outside of the lipid nanoparticle to guide it to specific cell types.

We’ve worked through a series of iterations to arrive at a targeted LNP chemistry that achieves delivery to T cells in mice. Our initial attempts barely achieved delivery of RNA to T cells, but over the course of a few months we were able to nearly match the current state-of-the-art.

With these tools in hand, we can now test whether the payloads emerging from our Discovery Engine are effective in the complex environment of an organism.

Measuring endothelial cell age

Our most recent therapeutic program is focused on restoring function in endothelial cells of the vasculature. Before we can discover payloads that make old endothelial cells act young, we have to develop experimental assays to measure impairments with age.

Endothelial cell aging is dramatic enough that we’ve been able to discover functional impairments in just our first few experiments. Old human endothelial cells appear to be less regenerative than young cells in a simple culture system. Older tissues tend to lose the vascular networks that carry nutrients and signals throughout the body in a process known as “rarefaction.” These tissue level impairments may be connected to a cellular level loss of function like we’ve measured.

We further tested the impact of age in an animal model of endothelial injury in the kidneys. We found that aged animals display far more damage than their young counterparts, suggesting the aged endothelium may be less resilient to damage as well.

These are only the first of many functional evaluations we’ll employ. Already, they provide us with an initial tool to evaluate the functional effects of reprogramming payloads we discover.

Join our team

We’re recruiting for talented scientists across various departments. If you’re excited about the possibility of bringing the first reprogramming medicines to patients, please reach out.

NewLimit is developing medicines that can restore youthful function in old cells through epigenetic reprogramming. Since our Series B, we’ve already seen significant advances in our lead programs that give us the confidence to start planning clinical studies in the near term.

At NewLimit, we are firm believers that a new class of longevity medicines can improve the lives of everyone on the planet, even those not diagnosed with classically defined disease. We appreciate each of these new capital partners joining to support our mission.

A few highlights:

• +2 prototype medicines with efficacy in preclinical models of liver disease

• +14 transcription factor sets that restore youthful function in T cells

• +8 transcription factor sets that restore youthful function in hepatocytes

• >4000 transcription factor sets tested across therapeutic areas

• 1.12X improvement in real discovery rates from reprogramming AI

• 0 → 1 lead payloads with translational chemistry

• 0 → 1 screens in endothelial cells

People

We’ve welcomed several new members to the NewLimit team.

• Greg Farber joined our Vascular team as a Scientist. Greg previously worked in the Regenerative Medicine team at Genentech, completed a postdoctoral fellowship with Li Qian at the University of North Carolina, and earned a PhD at Cornell.

• Quinn Lyon joined our Immunology team as a Senior Research Associate. Quinn previously worked at Deciduous Therapeutics.

• Robin Meyers joined our Predictive Modeling team as a Data Scientist. Robin previously completed a PhD at Stanford with Paul Khaveri exploring the epigenetic regulation of epithelial fate decisions.

• Daniella Rastelli joined our Metabolism team as a Senior Research Associate. Daniella previously worked on metabolic disease models at Takeda.

• Meghan Sedovy joined our Vascular team as a Senior Research Associate. Meghan previously completed a PhD at Virginia Tech.

• Benjamin Humphreys and Matthew Breyer have joined our Scientific Advisory Board to support our new Vascular program.

Constructing therapeutic molecules

NewLimit has been building prototype RNA medicines in our Metabolism program for nearly a year. Our earliest prototypes used lipid and RNA molecules that are effective for work in animal models, but preparing for human trials requires an even higher quality and performance standard.

We spent the past few months upgrading our early prototypes to use clinical-grade lipid chemistries. We initially encountered performance challenges when we simply swapped in the new chemistry for the old without iterating on our manufacturing methods. Our team took these challenges in stride and performed an optimization campaign that eventually achieved superior delivery efficiency to our prior prototypes.

This new chemistry worked so well in fact that we’ve switched to using it for all of our future work in the program. Over the coming months, we’ll similarly be engineering optimal RNA sequences for subsequent use in humans.

Inventing medicines is a game of serial prediction, a parlay bet on the outcome of a product once it reaches a human trial. The more similar a preclinical system is to the eventual medicine, the more likely those early results are to replicate in human patients. While difficult to implement at the earliest stages, we believe our investments in deploying clinical-grade chemistry early in our discovery process increase our chances of delivering products that provide real benefit for humans.

Selecting for resilience

NewLimit’s Discovery Engine screens measure both phenotypes and functions of old cells treated with reprogramming interventions. Our phenotypic screens ask whether reprogramming makes an old cell look like a young cell, whereas our functional screens ask whether reprogramming makes an old cell act like a young cell. For our Metabolism program focused on restoring hepatocyte function, we’ve historically measured the regenerative capacity of human hepatocytes in a humanized liver as our main functional read-out.

Last month, we introduced a second functional screening paradigm to measure the resilience of hepatocytes to injury from toxic diets. In this setting, we engineer each hepatocyte in the liver to harbor a different combination of transcription factors. We then reprogram the cells with a dose mimicking an RNA therapy, then challenge the liver with an alcohol diet.

Reprogramming payloads that make hepatocytes more resilient to alcohol toxicity become more abundant in the liver over time as cells with other payloads die off. The system allows us to directly measure how well reprogramming protects hepatocytes from damage in hundreds to thousands of payloads at a time.

To evaluate the performance of this system, we retested several reprogramming payloads that increase or decrease resilience in gold-standard preclinical models. We know the effects these payloads should have, so they serve as a set of positive controls that allow us to measure the success of the screen. We were excited to find that our positive controls were strongly enriched for their expected effects, arguing that the screening system itself is performant.

Already, these first screens have revealed new therapeutic payloads with efficacy in animal models. We look forward to the many discoveries to come.

Restoring function in the body’s transit system

In principle, we believe epigenetic reprogramming can restore youthful function in the majority of cell types within the human body. We’ve intentionally narrowed our focus to a small number that maximize the potential impact of these medicines in NewLimit’s early years. With the rapid progress of our lead program in hepatic metabolism, we believe it’s now time for us to expand the palette of cell types we study. We launched our third therapeutic program focused on restoring function in endothelial cells of the vasculature as a result.

Why endothelial cells?

Endothelial cells line the veins, arteries, and capillaries that allow blood to circulate to all the other tissues in your body. They are one of the few cell types that participates in the health and function of every other tissue. As we age, our endothelium atrophies and becomes pathologically permeable. Our aged tissues have inadequate access to the circulation and are exposed to undesirable signals that promote disease. Endothelial dysfunction contributes to aging pathology in the renal, cardiovascular, and cognitive systems among others. Restoring youthful function in endothelial cells may therefore offer diverse benefits.

Our Vascular team is initially focused on improving function in aged & diseased kidneys. There’s strong evidence that kidney endothelial cells lose function with age, and that restoring this function can provide benefit to patients with otherwise intractable and debilitating disease. Kidney function declines in most older patients, so we imagine that a medicine restoring youthful endothelial function could one day benefit nearly everyone over a certain age.

Each of our therapeutic programs begins by translating our Discovery Engine technology to the new cell type of interest. Here, we found that our existing Engine tools were performant in endothelial cells with zero modifications, demonstrating the generality of our approach.

As a result, we completed the first successful screen for our Vascular program just weeks after launching the program. We’re now heads down building functional assays & preclinical models to measure the performance of payloads we discover in these screens.

In silico reprogramming with real world feedback

Modern artificial intelligence systems are often trained using feedback from humans or verifiable rewards to improve the quality of generated samples. We face an analogous challenge for our in silico reprogramming models. Our models design a number of reprogramming payloads intended to achieve a desired cell state (e.g. young, but keeping the same cell type). To determine if these designs are effective, we need to verify their performance in the real-world through laboratory experimentation.

We recently executed the first large scale reprogramming screen with a new “epistatic” variant of our in silico reprogramming models. These models take as input a representation of the transcription factor genes we deliver to aged cells and predict as an output the expected effect on cell age and identity. We can then use these models to design experiments that test payloads most likely to achieve our desired cell states. If our models are effective, they should allow for us to make more discoveries per experiment by prioritizing the payloads to test more effectively than human alternatives.

Our improved models demonstrated just this result. A large screening experiment designed with these models yielded 12% more discoveries than our previous state-of-the-art model. This represents a rare case where improvements in the rapidly refined world of bits lead to more value in the high latency world of atoms. We hope to share more on the technical details behind the iteration soon.

These new experimental data allow us to build even more performant models, following the data scaling laws we’ve reported previously. We shared technical details behind an earlier version of our in silico reprogramming approach at ICML in July – check out the paper!

Elsewhere

We had the pleasure of talking about NewLimit in a few other venues over the past couple of months. Check out these conversations at the links below.

• Dwarkesh Podcast, Evolution designed us to die fast; why we can change that

• ICML 2025 GenBio, In silico design of epigenetic reprogramming payloads

• The Story Company, The Secret to Accelerating Longevity

Work with us

There are a finite number of problems where the outcome matters on the timescale of a century. Creating abundant intelligence, unlocking sources of energy too cheap to meter, and allowing humanity to explore the stars may be a few examples. We believe that expanding the number of healthy, happy years in each human life is one of these few, rare problems with enduring impact.

Making progress toward our mission is commensurately difficult, but offers the opportunity to perform the best work of a career. If you’re compelled to pursue goals with deep meaning and excel to the top of your craft, please consider joining our team.

newlimit.com/careers

To support the program, we’re excited to welcome Matthew Breyer and Benjamin Humphreys to our Scientific Advisory Board (SAB). As we navigated through the intellectual inception of our program, advice from both Matt & Ben was invaluable. We’re excited to work with each of them to develop the first reprogramming medicines for vascular diseases.

Subscribe now

About Matthew Breyer

Matt has a unique background spanning both academic medicine and pharmaceutical development. Matt began his career as a Professor of Medicine at Vanderbilt, followed by his transition into industry as the CSO of Biotherapeutics Discovery at Eli Lilly and later as a Distinguished Scientist at Janssen. In our first conversations with Matt, it became clear that he not only could work with us to imagine future therapies unlike those of today, but help us remain grounded in the pragmatic realities of pre-clinical development and eventual patient care.

See Matt’s public bio, borrowed below.

Dr. Breyer received his MD from Harvard in 1979. Following residency and nephrology training he served as Professor of Medicine in the division of Nephrology at Vanderbilt University Medical School from 1985-2007, where his work focused on causes of diabetic kidney disease and hypertension focusing on transgenic mouse models. As a result of his interest in developing treatments for diabetic nephropathy he moved to Eli Lilly in 2007 to oversee a program to develop bio-therapeutics for diabetic kidney disease. In 2009 he was appointed to the position of Chief Scientific Officer Lead Generation Biotechnology Discovery Research at Eli Lilly where he oversaw preclinical target identification and validation of biologics for chronic kidney, metabolic, neurologic, and oncologic diseases, and served until 2018. In April 2018 Dr. Breyer joined the Janssen Cardiovascular and Metabolism group in Boston where he guides discovery and development of new drugs for kidney disease.

About Benjamin Humphreys

Ben is not only a leading physician in nephrology at Washington University, but a successful biotech company-builder through his work cofounding Chinook Therapeutics (acquired by Novartis in 2023). We’ve long admired Ben’s work dissecting the basis of renal diseases with modern genomics methods, pushing forward the cellular and molecular understanding of the entire field. Our early conversations with Ben have helped guide our decisions for early discovery experiments all the way through eventual clinical plans.

See Ben’s public bio, borrowed below.

Dr. Benjamin Humphreys received his AB from Harvard college and his M.D. and Ph.D. degrees from Case Western Reserve University as part of the Medical Scientist Training Program. He trained in internal medicine at Massachusetts General Hospital and in nephrology at Brigham and Women’s Hospital and Massachusetts General Hospital. He was a member of the faculty at Harvard Medical School from 2005 – 2015, rising to the rank of Associate Professor. In 2015, Dr. Humphreys moved to Washington University in St. Louis, where he is Chief of the Division of Nephrology and is the Joseph Friedman Professor of Renal Diseases.

A few highlights from our output:

• +3 TF sets with pre-clinical efficacy in animal models of liver disease

• +11 TF sets that restore youthful function in aged T cells

• +>4000 TF sets tested across hepatocytes and T cells

• 100X increase in compute efficiency for reprogramming payload design models

• >60% increase in hit discoveries in active learning campaigns

Subscribe now

People

We’re excited to welcome several new members to the NewLimit team.

• Ron Hause is joining as our Head of Computational Sciences leading our Computational Biology and Predictive Modeling efforts. Ron previously served as SVP and Head of AI at Shape Therapeutics, helping bring RNA medicines from the whiteboard toward the clinic.
  
  Ron began his career in adoptive cell therapy, contributing to the development of Breyanzi and leading predictive modeling at Juno, Celgene, and BristolMyersSquibb. Ron's academic training began with a PhD in Genetics at the University of Chicago and culminated in a Damon Runyon postdoctoral fellowship under genomics pioneer Jay Shendure at the University of Washington. He brings over 15 years of experience at the intersection of statistical genomics, machine learning, and therapeutic development to help scale our computational capabilities.
  
  We’re excited for Ron to help us bridge the gap between early discovery and therapeutic development.

• Ryan Theisen joined our team earlier this year as a Senior Machine Learning Scientist. Ryan previously developed generative models for small molecule design at Harmonic Discovery. In his academic training, Ryan performed celebrated research on model ensembling and completed a PhD in Statistics at UC Berkeley.

• Michael Herschl joined our Functional Genomics team as a Scientist. Michael recently completed Ph.D. training at UC Berkeley and the Arc Institute where he developed pooled screening systems for epigenetic editing.

• Nehal Hedge joined our Functional Genomics team as a Research Associate. Nehal started her career at Pragma Bio and completed her Bachelor’s degree at the University of Washington.

Restoring youthful function in multiple dimensions

The premise of NewLimit is that aging leads to a decline in cell function that contributes to diverse pathologies. If we’re able to restore youthful function in old cells, we imagine this could likewise help treat multiple diseases and quality of life impairments that arise with age.

We previously discovered reprogramming payloads that restored youthful gene expression in old hepatocytes and formulated these payloads into prototype medicines. Earlier this year, we demonstrated that our prototype medicines could rescue function in one animal model of liver disease. Prototype medicines that reverse cell age should be effective across more than just one of these functional dimensions.

This summer, we explored whether these prototype medicines could restore youthful function in multiple disease models. If a payload can restore more than one youthful function, it suggests that underlying features of age have been reversed. We were excited to discover that a prototype medicine emerging from our Discovery Engine screens (M0004) not only improves hepatocyte regeneration, but also appears to restore youthful resilience to damage from alcohol.

These data are still preliminary, but deepen our conviction that it’s possible to restore youthful function across multiple dimensions.

Generative design of reprogramming payloads

There are ~10^16 possible combinations of TFs that might reprogram cell age. This represents ~10,000X the number of stars in the Milky Way galaxy. No matter how sophisticated our experimental systems become, we will never test all of them.

From our earliest days, NewLimit has been developing an artificial intelligence (AI) system to search this space by performing reprogramming experiments in silico. Earlier this year, we demonstrated that our in silico reprogramming models could predict reprogramming effects 2X better than the state-of-the-art in the broader field. These models allowed us to answer the question: What will happen if we reprogram old cells with TFs X, Y, and Z?

Designing the optimal payload to reprogram cell age requires us to ask a related, but different question: What combination of TFs is most likely to reverse cell age, while preserving cell type? It turns out that mathematically, this latter question is a generative problem. Exhaustively predicting the effect of all possible reprogramming payloads with our models is possible, but quite expensive – >$1M! We therefore need an efficient algorithm to design payloads.

Our Predictive Modeling team overcame this bottleneck by developing a Bayesian generative process that allows us to design optimal payloads with >100X less compute than before. This approach not only allows us to efficiently deploy our models into our production Discovery Engine experiments, but also enables us to search even larger hypothesis spaces in the future, including the design of synthetic TFs.

The true test of these methods is whether they can accelerate our reprogramming discoveries by recommending new hypotheses to test after each screening experiment. This is akin to an active learning or Bayesian optimization setting. We’ve now demonstrated that using our new payload design algorithms can increase our discovery rate by >60% in a simulated campaign.

Production-grade reprogramming screens

The data corpus that fuels our in silico reprogramming models is the result of relentless experimental progress in our Discovery Engine. When we started two and a half years ago, every element of our technology had to be built from scratch. Nothing was optimized, and we had yet to even determine which parameters were most critical to tune. We couldn’t execute as many experiments as we desired, and we discovered many quality control challenges.

Each experiment has yielded lessons that allowed us to improve on both of these metrics. Quarter over quarter, we’ve continued to increase the number of screens we can execute. Alongside an increase in throughput, we’ve instrumented our process end-to-end to reduce error rates. There was no singular update that explains our progress in these dimensions, but rather a process of continual improvement that led to countless small changes. That process itself was only possible due to the efforts and ingenuity of our team.

This summer, we’re proud to complete the quarter with our highest throughput and lowest quality control failure rate on record. We’re able to turn investments of time and resources into more discoveries than ever.

Our most recent screens have contributed thousands of new reprogramming payloads to our growing data corpus. We’ve now tested >700X as many payloads as the rest of the field combined to our knowledge.

We believe the scale of this data corpus is one of the key reasons we’ve been able to build performant in silico reprogramming models. Since our initial report of data scaling laws for reprogramming, we’ve found that scaling continues in the larger data regime. These results give us confidence that our investments in experimental throughput and quality will continue to yield dividends.

We’re at the beginning

All of us at NewLimit are excited by the data that now readily emerge from our models and laboratory. Alongside our recent fundraising, we’re expanding to build reprogramming medicines across even more therapeutic areas. Nonetheless, we’re still at the beginning of the story. Our impact will only grow from here.

If you’re motivated to build medicines that add healthy years to every human life, please consider joining our team for the upcoming chapters.

Subscribe now

NewLimit is building medicines to treat aging, the meta driver of nearly every major human disease. Once thought to be irreversible, the emerging science of epigenetic reprogramming has shown that aging is in fact malleable. This presents a therapeutic opportunity that is 100X larger than any biotech product that has come before.

NewLimit’s approach is conceptually straightforward – we’ve built technology to run the largest experiments in the field, and we use the resulting data to train frontier AI models that ensure we only test the most promising potential medicines. Our bet is that this combination of AI and scaled genomics can unlock medicines that give each of us more healthy years. We have a stepwise, capital efficient plan to get there.

We're deeply appreciative of investors who share our vision that addressing aging biology represents one of the most consequential projects of humanity. We hope our progress will help inspire broader interest and investment in aging therapeutics.

Our story is just getting started. If you want to be a part of it, join us at newlimit.com/careers.

You can watch a recording our presentation below where we reflect on our progress over the past few years and outline our goals for those to come.

As always, if you’re excited by our mission, please consider applying to join NewLimit.

A few highlights from the first months of winter:

• +3 TF sets with pre-clinical efficacy in animal models of liver disease

• +3 TF sets that restore youthful function in aged T cells

• +10 TF sets that make old T cells look young based on gene expression

• 20X increase in humanized liver screening bandwidth

• 2X better performance for in silico reprogramming than the best baseline

People

We welcomed two new team members at the beginning of the year:

• Kelly Chu joined as an Operations Specialist to upgrade our laboratory & office operations.

• Julianne Obiniana joined as an Operations Specialist to support our G&A functions.

Restoring youthful function in hepatocytes

At the end of 2024, we completed our first reprogramming screens in humanized livers and discovered transcription factor (TF) sets that made old hepatocytes both look and act young. We formulated the top hit into a prototype LNP-mRNA medicine and demonstrated that it restored youthful function in a pre-clinical model of liver disease.

In the first months of 2025, we’ve repeated that process and discovered 3 more TF sets that restore youthful function when delivered with prototype LNP-mRNA medicines. We refer to these prototype medicines as leads.

Our first lead rescued the ability of old hepatocytes to regenerate the liver after damage. Youthful hepatocytes are not only more regenerative once damage has occurred, but they’re also more resilient and less likely to die in the face of damage in the first place. This month, we built prototype medicines based on 2 unique TF sets that make old hepatocytes more resilient, mimicking the performance of younger cells.

We demonstrated improved resilience in an ethanol injury mouse model. This model mimics the damage caused by alcohol in human liver disease. In this context, we can measure the amount of damage that hepatocytes experience using biomarkers of hepatocyte death in the blood.

After treatment with our prototype reprogramming medicines, old livers show a near youthful level of resilience to injury. To our knowledge, this is the first time that a prototype partial reprogramming medicine has been shown to treat an alcohol-related liver disease model.

We took a gamble when we tested one of these successful TF sets. We were excited because it made old hepatocytes look young based on gene expression profiles in our pooled screens, but we didn’t have much evidence that it could improve youthful function. This payload hadn’t been studied much at all outside NewLimit. These functional results suggest that making old hepatocytes look young is a reasonable way to discover reprogramming payloads that make old hepatocytes act young. It’s hard to imagine better motivation to scale our screens!

We have now demonstrated 4 separate prototype reprogramming medicines that treat disease in pre-clinical models. Prior to our initial results in late 2024, this had never been shown before.

Restoring youthful function in T cells

T cells are key orchestrators of the immune system. In one of their many roles, CD8 T cells kill “target” cells that present foreign antigens like virally-infected cells or cancer cells. We’ve found that young CD8 T cells are much better at killing target cells than their old counterparts. This result suggests that old T cells may be less effective at clearing infections and cancers.

Last month, we discovered 3 TF sets that restored youthful killing activity in old CD8 T cells. We believe this is the first demonstration that partial reprogramming can rescue human CD8 T cell function. Similar to our functional experiments in hepatocytes, we performed these validations with a prototype mRNA medicine, mimicking a realistic therapeutic modality.

The raw data from this experiment is shown in the timelapse microscopy videos below. Here, the target cells are engineered to express a red fluorescent protein. If the T cells are functioning well, the red cells don’t grow and there is less red in the frame.

On the left, a culture without T cells allows for rapid target cell growth. Old T cells treated with a negative control mRNA in the center can modestly prevent target cell growth. By contrast, old T cells reprogrammed with our top lead I0001 on the right are able to restrict growth similar to young T cells (less red in the frame).

We’re excited to carry these results forward into pre-clinical models.

High-bandwidth humanized liver screens

We completed the world’s first reprogramming screen in humanized livers at the end of 2024. This system is powerful because it allows us to simultaneously discover TF sets that make old human hepatocytes look young based on gene expression and act young based on regenerative potential.

In January and February, we doubled down on our investment in these tools and improved the technology to increase screening bandwidth by 20X. This means that for every humanized liver we build, we can now test 20X as many TF sets.

There was no singular trick that unlocked this improvement. Rather, a combination of improved molecular engineering, animal model development, and genomics workflows came together to yield an order of magnitude increase in performance. We’ve already completed our first screens and launched several more using this improved technology.

Performant in silico reprogramming

No matter how many TF sets we test in the laboratory, we’ll never be able to exhaustively search the space of all possible reprogramming payloads. The number of possible TF sets we might target is roughly 10¹⁶, or about 10,000X the number of stars in the Milky Way galaxy. Experimental efficiency is critical but insufficient to explore this space. We also need methods to help us prioritize which experiments to run in the first place.

To expand our search space, we’ve built AI models that can perform reprogramming experiments in silico. These models take in a representation of an old cell state and a set of TFs and predict the effect of partial reprogramming on cell age and type. We use these predictions in the world of bits to prioritize our work in the world of atoms.

This month, we demonstrated that our models are superior to the best baseline method in the broader field of genetic perturbation prediction. Our models can now explain more than half of the variation in cell age effects for unseen TF sets.

Our model’s strong performance is enabled by two main factors: (1) our proprietary data corpus and (2) recent advances in foundational models of biology. We’ve observed a data scaling law in our in silico reprogramming models where performance increases as a function of our training data scale, similar to results in natural language and computer vision. As our data set grows, our models get better, and as our models get better, the marginal data point we collect increases in value.

Foundational models of biology allow our in silico reprogramming models to start from rich representations of TFs that emerge from their DNA and protein sequences, instead of learning all of TF biology de novo. This means that our models can make reasonable predictions for new TF sets we’ve never tested before based on the TF sets that we have already tested.

We’ve already incorporated these models into our production screening process, and we look forward to sharing the results of our active learning campaigns soon.

Our team is growing

We are always recruiting talented scientists, engineers, and operators. If you’re excited by our mission to add healthy, happy years to every human life, please reach out about our open roles.

1. >10X increase in the throughput of our Discovery Engine

2. Discovered TF sets that restore youthful gene expression while preserving cell type

3. Demonstrated that partial reprogramming with a drug-like molecule can restore youthful function in an animal model of liver disease

Along the way, we answered a series of open scientific questions that stood between us and reprogramming medicines. We believe that several of the risks present at our founding were overcome in the past 12 months.

Subscribe now

Engine Cadence

NewLimit has built a set of technologies we refer to as our Discovery Engine to discover partial reprogramming payloads. Concretely, our Engine allows us to measure the effect of briefly activating a set of transcription factors (TFs) on the genes a cell is expressing and the functions it can perform.

In a single experiment, our Engine measures whether any of thousands of TF sets can make an old cell look younger – express genes closer to a young cell. For the most promising TF sets, we can then test if reprogramming makes old cells act younger – performing functions more like their youthful counterparts.

This year, we scaled the throughput of the Engine >10X along several axes. Our progress is best illustrated numerically:

The throughput of our Engine is the first-order input to the pace of our therapeutic discoveries. Outside NewLimit, we’re aware of <20 TF sets that have been tested for their effect on cell age. We now routinely run individual experiments that generate 50X as much data. Cumulatively, we’ve tested more than 500X as many TF sets as had been previously reported.

These screens resulted in our first promising discoveries. We found 100+ TF sets that restore youthful gene expression in old, human cells, and 5 sets that restore a youthful function.

Restoring youthful expression & preserving cell type

NewLimit was originally inspired by remarkable results from pluripotent reprogramming experiments. By activating just four TFs – the Yamanaka Factors – old cells can be reprogrammed back to a youthful state. These cells can give rise to a young animal with a normal lifespan. In old animals, transient activation of these factors can restore youthful gene expression and function across several tissues.

These discoveries are an existence proof that partial reprogramming can restore youthful function. Unfortunately, the Yamanaka Factors also suppress adult cell type programs, potentially turning adult cells into embryonic cells. This introduces a risk of neoplasia and limits efficacy, making it challenging to use them as medicines.

When we founded the company, we wondered if we could disentangle these two effects. Can we find sets of TFs that reprogram cell age, without reprogramming cell type? If so, medicines become tractable.

In 2024, we’ve found sets of transcription factors (TFs) that restore youthful gene expression, without abolishing cell type programs in both T cells (a key immune system cell) and hepatocytes (liver cells). These discoveries stand in contrast to the Yamanaka Factors which make old cells look younger at the expense of compromising their cell types.

Restoring youthful expression is compelling, but therapeutics may require these changes to be durable. A medicine that can be dosed every few weeks is more useful than one that must be dosed every day. We also discovered TF sets that restored youthful expression long after the TFs were turned off, mimicking durability after the withdrawal of a drug.

These results address two of the core scientific risks that were present at NewLimit’s founding. Youthful expression can be restored without abolishing cell type, and these effects can endure over time. There are many miles to go before we have medicines in hand, but these data suggest that nature is amenable to their existence.

Restoring youthful function with therapeutic molecules

Function is the final boss at NewLimit. We ultimately care most about making old cells act like young cells, restoring youthful functions that will treat aging & disease in patients. At the start of the year, we had reason to believe that we could build reprogramming medicines to achieve this goal, but no actual data.

In early winter, we developed the capabilities to produce lipid nanoparticle-mRNA formulations. These molecules are capable of briefly activating TF sets in hepatocytes in the liver. Importantly, they can also be turned into medicines. Pioneering LNP-RNA medicines are already approved in the US, including medicines that target hepatocytes in the liver and the COVID vaccines.

If partial reprogramming with these molecules can restore function, it offers a path to building reprogramming medicines.

We ran our first in vivo screens in human hepatocytes late in the autumn. To our surprise, even these early screens revealed multiple TF sets that restore youthful gene expression and function. Seeing functional improvements in a screen is encouraging, but we next wanted to confirm that these effects were robust.

We formulated one of these TF sets as an LNP-mRNA and used it to treat a model of liver injury in old mice. Across several experiments, we found that partial reprogramming with these drug-like formulations was sufficient to restore youthful function in these old cells. The regenerative capacity that was lost with age remained latent and reemerged following reprogramming.

This model of age-related disease is only the first of many we plan to employ, but the results are nonetheless a promising sign that the medicines we envision are possible.

Technical innovations

Discoveries are downstream of the technologies that enable them. We’re especially proud of four technological innovations in our Engine last year.

In silico reprogramming performance scales as a function of data

Modern artificial intelligence systems frequently exhibit scaling phenomena – performance increases as a function of training data size, diversity, and compute.

We use in silico reprogramming (ISR) models to predict the effect of partial reprogramming with a given set of TFs, then prioritize our experiments using these predictions. These models are critical because the search space of TF sets is massive – 10¹⁶ combinations are possible! We can never search the entire space experimentally. In silico predictions allow us to target our experiments to only the most promising hypotheses.

This year, we showed for the first time that the performance of these predictions increases as a function of data scale. This is one of a handful of examples suggesting that scaling phenomena may exist in learned representations of life, in addition to the traditional artificial intelligence domains of computer vision and natural language.

We appear to still be in the linear regime of data scaling, suggesting that our models will continue to improve as we rev the Engine in 2025.

In vivo screening in human hepatocytes

In 2024, we initiated our Metabolism program focused on restoring youthful function in hepatocytes. One of the first challenges we encountered is that performing partial reprogramming screens at scale is challenging in this cell type.

Hepatocytes are not amenable to the in vitro culture techniques used in most screening systems. To overcome this challenge, we developed an in vivo screening technology for our program. This approach takes advantage of humanized liver mouse models to measure reprogramming effects in human cells inside a real, functioning liver.

To our knowledge, this is the first time a human hepatocyte screen has been executed in vivo. We believe the data emerging from this platform are more likely to reflect the biology of human aging & disease than more common animal model systems.

Functional assays of human cell age

Ultimately, NewLimit’s goal is to make old cells act young, not just look young. In order to achieve this goal, we first need to develop tools that can measure the functional defects that emerge in old cells.

We developed 11 new functional assays last year that discriminate between young and old cells, 5 in T cells and 6 in hepatocytes. While it may seem obvious that old cells are less functional than their young counterparts, measuring this rigorously is a surprisingly understudied area. We believe we now have the most robust assay suite for cell age that’s been developed for our therapeutic areas.

Combinatorial indexing for combinatorial screens

NewLimit is fairly unique in that we’re searching for therapeutic payloads that contain multiple genes. Screening for combinations introduces several technical challenges. One of them is quite simple – we need to be able to detect all of the TFs we’ve delivered to a given cell in order to measure their effect.

We developed an initial method to perform combinatorial TF screens in 2023. This year, we developed a new genomics method called pxbc-seq that builds upon combinatorial indexing methods to cut the cost of some of our screens by >100X. This technology allows us to perform selection-based screens cheaper than ever before. Already, we’ve deployed it to test >20X more TF sets in a functional assay than we could test with a traditional approach.

2025

The story of NewLimit’s past two years can be summarized concisely.

• 2023: We built the technology to discover partial reprogramming payloads in human cells.

• 2024: We discovered payloads that reprogram cell age while preserving cell type. We showed that reprogramming medicines are possible.

• In 2025, we hope to progress some of these discoveries from research into the early stages of therapeutic development, laying the groundwork to bring medicines to patients.

There is a long road ahead before those medicines reach humans, but we’ve made more progress than we expected to date. We feel more confident today than ever that reprogramming medicines will have a marked impact on human healthspan.

Join us

The achievements we can claim thus far are a testament to the talent and effort of the team we’ve built. We continue to hire across multiple functions. If you’re excited by our mission, please reach out.

After a minute-long celebration, we put our heads back down. It makes me proud that as we pass this annular boundary, we have more firsts to celebrate on a grander scale.

A few highlights from the first months of autumn:

• +2779 TF sets tested in primary T cells

• +51 TF sets that reverse the phenotypic age of T cells

• +422 TF sets tested in T cell functional assays

• 0 → 70 TF sets tested in human hepatocytes

• 5 → 7 functional assays for hepatocyte age

• 0 → 80% efficiency for in-house LNP delivery to hepatocytes

Functional screens in primary T cells

Our Immunology program has to date discovered rejuvenating TF sets through large scale phenotypic screens, then tested these TF sets in functional assays downstream. We continued to expand our corpus over the past two months and reached an all-time-high number of TF sets tested for the quarter (2779).

We also debuted our first functional pooled screening method for the Immunology program. This approach combines a mechanism to separate functional and dysfunctional cell populations developed by our Immunology team and a low-cost, high-throughput cell profiling chemistry developed by our Read team. We defined these functional & dysfunctional populations using one of our age-discriminating T cell functional assays.

In each experiment, we deliver a pool of TFs to old T cells and each cell acquires a random subset of TFs, just like in our phenotypic screens. We then transiently activate TFs to perform partial reprogramming and segregate the cells based on their functionality. Finally, we measure which TF sets are more abundant in the functional population relative to the dysfunctional population. If a TF set appears more often in the functional cells, it suggests that reprogramming with this set improved cell function.

We found that this new assay was able to test >400 TF sets in about the same time that it previously took us to test 20, for a >20X improvement in throughput. Results from this new method were highly correlated with our previous low-throughput experiments, giving us confidence that not only do these results reproduce, but also that our new assay is accurate.

In vivo reprogramming screens in human hepatocytes

Our Metabolism program is inventing reprogramming medicines to restore youthful function in aged hepatocytes. We hope these medicines will eventually treat common diseases of aging that affect us all. The number of possible combinations of transcription factors (TFs) that might compose our medicines is massive. Our Write (Functional Genomics) & Metabolism teams have developed a way to onboard human hepatocytes to our Discovery Engine to navigate this hypothesis space at scale.

Traditionally, therapeutics companies might screen for ways to improve hepatocyte function using cancer cells or primary cells grown on plastic dishes. There are a few problems with this approach. Cancer cells are quite different from normal human biology, and primary human hepatocytes really don’t grow well in a dish, confounding results. Experiments in these settings don’t have a great track record of translating into humans.

We believe that the results of our Engine are only as reliable as the biology of our model systems. The best model of human hepatocyte biology to our knowledge is the chimeric humanized liver system developed by our SAB member Markus Grompe. In this system, human hepatocytes are engrafted into the liver of a genetically-engineered mouse and begin to regenerate, eventually resulting in a largely human tissue within a living animal.

Throughout the summer, our team leveraged the core technology of our Discovery Engine to build the world’s first pooled system in humanized livers to our knowledge. Our technology allows us to grow humanized livers where each human hepatocyte receives a transient pulse of TF mRNA from a random set of TFs within a pool. Across this “mosaic” liver, hundreds to thousands of unique TF sets are represented.

In a single liver, we can test whether each of these TF sets make old hepatocytes look young based on gene expression and act young based on their regenerative performance. We know that old hepatocytes are much less regenerative than young hepatocytes, so this functional result is a useful proxy for the restoration of youthful function.

This month, we read-out the first results from our reprogrammed chimeras. Already in the first 70 TF sets we’ve tested, we’re seeing promising signals. We have several more of these experiments in flight now, and hope to increase the number of sets we’ve tested by several fold before the year is out.

Pre-clinical models of hepatocyte age

Our Discovery Engine screens provide us the first signals to prioritize TF sets that might rejuvenate aged cells. Ultimately, we will evaluate these TF sets in a series of scalable functional assays and traditional pre-clinical models.

This September, we measured the impact of hepatocyte age across two preclinical animal models of age-related liver disease. Each of these models mimics the liver damage that can occur from modern diets. We found in both cases that aged hepatocytes were significantly more vulnerable to this damage than young cells, suggesting that rejuvenating hepatocytes may benefit patients with similar diseases.

In total, we’ve now found that aged hepatocytes perform worse than young hepatocytes in 7 distinct functional assays and pre-clinical models. With each of these data points, we’ve grown more confident that rejuvenating hepatocytes would provide meaningful benefit to aged patients.

In vivo delivery of reprogramming factors with therapeutic chemistry

Discovering TF sets that restore youthful function is only the first step in inventing a reprogramming medicine. Once we’ve discovered a payload of TFs, we next need to formulate it into a drug-like molecule. NewLimit is developing our initial medicines using LNP-mRNA technology, similar to the COVID-19 vaccines.

This October, we took the first steps to start preparing our own LNP-mRNA formulations for in vivo testing. These first molecules are only intended for research at the moment, but bringing this capability in-house has allowed us to quickly progress initial hits from our Discovery Engine to in vivo functional testing. We’re proud of our team members for quickly matching best-in-class performance for hepatocyte delivery with our internal tools.

An idea factory

History's great laboratories were designed with intention. Be it the long hallways of Bell Labs, the small cramped offices of the Laboratory of Molecular Biology, or the desks interspersed amongst the prototyping floor in the Skunk Works, form begets function within scientific institutions, as in biology. 

NewLimit’s headquarters has always been amply provisioned for our work, but from the first days in the lab, we’ve had designs on modifying it to match our culture. In October, we finally completed a remodel and moved our team into refreshed space.

We stole ideas for the layout liberally. 

Our new main office is elongated, mimicking Bell Lab’s classic design. We replaced every wall that wasn’t load bearing with glass, ensuring that we feel connected to our colleagues even when we’re doing deep work. Following Kelly Johnson’s guidance, we also replaced the wall dividing our offices from our labs with a floor to ceiling window, ensuring that the ground-truth of experimental work on the “shop floor” is always close at hand. We couldn’t be happier with how it turned out.

A few highlights from July & August:

• +2500 TF sets tested

• +20 TF sets that reverse the phenotypic age of T cells

• 2X acceleration of reprogramming library Build cycles

• 0 -> 3 hits with preliminary functional effects in primary T cells

• +3 functional assays for hepatocyte cell age

• +4 team members

People

We had a rapid few months of growth across several key roles. We’re excited for our new colleagues to join the team:

• Adam Ferguson is joining us from Ginkgo as a Senior Automation Engineer on our Read team.

• Kate Franz is joining us from Verily as a Senior Scientist on our Immunology team.

• Catherine O’Hare is joining us from GlaxoSmithKlein as our Head of Operations.

• Rowena Suriben is joining us from Kinnate Biopharma as an Associate Director to lead our Metabolism team.

Early functional effects of partial reprogramming

Our Discovery efforts are ultimately all in service of developing medicines that improve the function of old cells and treat disease in patients. For our T cell program, we’ve now discovered dozens of transcription factor (TF) sets that make old cells look young based on our phenotypic screening assays. Until this summer, we had yet to test whether any of these sets make old T cells act young, improving their function to youthful levels.

Our Immunology team has now developed a series of functional assays that can discriminate between young and old T cells. With these assays in hand, we began testing whether partial reprogramming with our top phenotypic hits could improve T cell function on one of these assays in July. 

We found preliminary evidence that 3 of the phenotypic hits we tested improved the performance of old T cells in this assay, suggesting that at least this particular function is amenable to reprogramming. These results represent our first complete cycle of phenotypic discovery followed by functional validation, and we’ll be radically increasing the throughput of these validations going forward.

We believe it’s important that our eventual medicines improve T cell function across multiple axes. These results are encouraging, but only the beginning of our validation process that will feed into therapeutic development.

Scaling the data corpus

Alongside our first functional validations, we’ve continued to run our Discovery Engine at a steady clip. We tested another 2500+ TF sets for their effect on cell age in old human T cells across July & August, including our largest screen yet (and the largest in the world that we know of). 

From these experiments, we discovered another 20+ TF sets that reverse the phenotypic age of T cells through partial reprogramming. Each of these new hits provides both a direct substrate for our validation assays and contributes to the development of our in silico reprogramming models by adding to our data corpus. With each new screen, we discover additional axes of cell state that are amenable to partial reprogramming. 

We hope that these diverse data will allow us to design better experiments in the world of bits, then discover more hits per dollar in the world of atoms in the near future.

Accelerating Discovery Engine screens

Each of our Discovery Engine screens requires us to build a set of nucleic acid reagents to activate TFs in old cells, then detect which TFs are present to enable pooled experiments. This Build process is one of the most labor intensive steps of each screen and increases our turn-around time from one experiment to the next. We’re firm believers that reducing cycle times in our process will allow us to learn and improve that process more rapidly, ultimately yielding more discoveries and downstream assets.

Our Write team has been working for the past quarter to develop a new molecular biology process that cuts our Build time and costs in half. We developed a “two-step” Build process in the early days of NewLimit to ensure high fidelity – almost all of the reagents we start building complete successfully. However, the total amount of effort required was roughly 2X the theoretical minimum. Through a clever series of molecular tricks encoded in base pairs, our team was able to achieve a “one-step” process that preserves nearly the same fidelity as our original version.

In practice, this improvement cuts the turn-around time and the cost of our Build processes by roughly 2X. Through a series of improvements like this, we hope to continually improve the performance of our Engine.

Measuring functional age in human hepatocytes

The clinical evidence that human livers decline in function with age is quite clear, but measuring these defects in human hepatocytes in the laboratory is surprisingly understudied. Our Metabolism team previously developed an assay to discriminate young and old human cells based on their metabolism, and over the past two months we’ve also built a system to discriminate human cells based on their regenerative potential.

Hepatocyte regeneration is an important function that allows the liver to tolerate damage from environmental insults like a Western diet. We’ve found that regeneration declines with age in old mice, and our team developed a clever functional assay to demonstrate that the same is true in human cells.

We transplanted young and old human hepatocytes into humanized liver mice, then measured how well they regenerated. The humanized liver system, originally developed by our SAB member Markus Grompe, allows human cells to engraft in a “sick” mouse liver and slowly regenerate it over time until it’s mostly human. The more regenerative a cell is, the more abundant its progeny will be in the resulting humanized liver.

We found that young human hepatocytes are consistently more regenerative than old human hepatocytes, by a factor of >6X. The performance of human cells in this assay is sufficient to discriminate the age of the donor with strong performance. We can take a reasonable guess at a donor’s birthday based simply on how their hepatocytes perform in our laboratory. 

It’s worth emphasizing that this assay is cell-intrinsic and long-term. Even when surrounded by the exact environment over a long time period, old human hepatocytes maintain intrinsic defects relative to their young counterparts. Hepatocyte regeneration is also associated with positive outcomes for liver disease patients, supporting the clinical relevance of this cell function.

Our Metabolism team is now hard at work deploying this new tool to discover rejuvenating factors.

Come join us

NewLimit is always recruiting talented scientists, engineers, and operators. Right now, we have exciting opportunities for a Lipid Nanoparticle Delivery Scientist to build out our delivery technology and for a Machine Learning Scientist to own our in silico phenotyping suite.

Please reach out if you or any colleagues would be a good fit!

Through the spring and early summer, we’ve focused on developing functional assays for Immunology & Metabolism programs, improving the efficiency of our Discovery Engine screens, and revving the Engine to construct our reprogramming data corpus.

A few highlights by the numbers, quarter-over-quarter:

• 2X reduction in the cost per transcription factor (TF) set tested

• 1.7X the number of TF sets tested

• 1.5X the number of TF sets that reverse the phenotypic age of T cells

• +4 functional assays for hepatocyte cell age

• +1 functional assay for T cell age

• +3 team members

• +1 SAB member

People

We’ve also been excited to welcome new members to our team:

• Yu-Hsuan Lin joined as a Scientist II on our Metabolism team. Yu-Hsuan previously trained with our SAB member Hao Zhu and developed CAR-NK therapies at Nkarta.

• James Briggs joined as a Senior Scientist on our Write (Epigenetic Editing) team. James previously trained with Allon Klein and Marc Kirschner at Harvard and Feng Zheng at the Broad.

• Fatima Rizvi joined as a Scientist on our Metabolism team. Fatima previously trained with Valerie Gouon-Evans at Boston University.

Improving the efficiency of our Discovery Engine

Our Discovery Engine is designed to test partial reprogramming interventions at scale. Our goal is to explore as many hypotheses as possible on a given budget. If we can reduce the cost of testing each hypothesis, we can test more of them, increasing our chances of finding reprogramming factors that reverse cell age.

Over the past few months, our Write (Epigenetic Editing) and Read (Single Cell Genomics) teams developed two new approaches to increase the number of interventions we can test per dollar. First, our Write team developed a strategy to reprogram and purify cells more efficiently in our screens, decreasing the number of cells we need to profile to test each set of factors by 1.4X. Our Read team then developed a new single cell genomics chemistry that decreases the cost of each cell profile 1.4X. 

In our most efficient Engine experiments, these changes combined reduced the cost per TF set tested by 2X.

Reversing phenotypic age in primary human T cells

Our Discovery Engine screens are only as impactful as the biology they reveal. The primary question we ask in each of our screens is whether we’ve made an old cell look young based on its genomic profile. Concretely, we use machine learning models that predict cell age from cell genomics profiles to measure whether a set of transcription factors (TFs) can shift an old cell to a younger looking state.

Over the past few months, we 1.5X’d the number of TF sets we’ve discovered that can reverse phenotypic age in primary human T cells. Among these hits, we’ve found that many TFs outside the Yamanaka Factor set can shift phenotypic age. Cell age reprogramming appears to be broader than just the first factors that the field discovered.

We also found that many of these TF sets can reprogram cell age durably, even after the TFs themselves have been turned off. There are still many questions to be answered, but these early results suggest that partial reprogramming for a brief time may lead to lasting changes in cell age. Check out our Progress Update video for a deep dive on a subset of these results.

Functional assays for T cell age

While reversing phenotypic age is exciting, reversing functional age is our ultimate goal. We want our reprogramming interventions to make old cells act young, not just look young.

Each of our Therapeutic teams are developing assays to measure the functional impact of reprogramming therapies on cell age. The first step in this journey is developing assays that can discriminate between young and old cells in the first place.

We previously reported on a set of functional assays that can discriminate between young and old human T cells. We’re excited to share some early data from one of these assays alongside our live Progress Update.

T cells are distributed across a range of distinct functional states, each contributing to different features of the immune system. Our team discovered that young and old T cells adopt different functional states in one of our assays, and this distribution of states is sufficient to reliably determine if a tube of T cells came from a young or old person. 

That final point is worth emphasizing – the function of young and old T cells is distinct enough that you can reliably predict a person’s age just by measuring how their T cells perform in the laboratory.

Over the past two months, we’ve found another functional assay for T cell age to add to our repertoire and look forward to sharing results in the future.

Functional assays for hepatocyte age

In addition to T cells, we’re developing reprogramming medicines to reverse cell age in hepatocytes. We’ve been building our Metabolism program throughout the spring, and we’re excited to share that our team has already developed our first set of functional assays for mouse and human hepatocyte age.

We found that young hepatocytes are significantly more regenerative than old cells, and this increased regenerative potential can protect young animals from insults that are debilitating for old animals. We also found metabolic changes that are sufficient to discriminate young and old hepatocytes in humans. Similar to our T cell assays, these differences are sufficient to predict the age of a human donor based only on how their hepatocytes perform in the laboratory.

These functional assays form a critical component in our Discovery Engine and their development represents a meaningful milestone for our newest Therapeutic team.

We’re looking for builders

We continue to expand our team across all levels of seniority. More so than any specific technical background, we’re excited to work with talented scientists, engineers, and operators who are passionate about our ambitious mission.

If you’re excited about our work, please reach out.

We’re excited to announce that Markus Grompe is joining our Scientific Advisory Board to support our Metabolism program.

Markus’ laboratory is responsible for much of our collective knowledge about human hepatocyte biology. Among other notable accomplishments, they developed a system to build models of a human liver in mice with a series of clever genetic tricks. These tools allow our scientists to study the effects of partial reprogramming in real, human cells, rather than cells from other species with quite different biology. Alongside his experience in the laboratory, Markus is an experienced company builder through his work at Yecuris and Ambys Medicines.

Markus has inspired and helped advise our Metabolism program from the earliest days. We look forward to working with him for years to come.

About Markus Grompe, MD

See Markus Grompe’s profile from Oregon Health & Sciences University and Wikipedia. Paraphrased below.

Dr. Grompe is a Professor at Oregon Health and Sciences University, the director of the Oregon Stem Cell Center, and the holder of the Ray Hickey Chair and the Director of the Papé Family Pediatric Research Institute. The Grompe lab is focused on developing cell and gene therapies for genetic disorders, especially metabolic disorders of the liver. Among other accomplishments, the Grompe lab developed the first mouse models that allow for competitive repopulation of hepatocytes and further developed the first humanized liver models.

Dr. Grompe has also served as the co-founder and Chief Scientific Officer of Ambys Medicines and the founder of Yecuris.

Dr. Grompe received his M.D. in 1982 at the University of Ulm Medical School in Germany. He completed training at OHSU and Baylor College of Medicine focused on gene therapy for inherited diseases, particularly liver diseases.

We’ll be hosting our second annual Progress Update live event on May 23rd, 2024 in South San Francisco.

We’ll be livestreaming the update, and also welcoming guests to join us on site. If you’re interested in attending in person, apply through the form below.

Apply to join us live!

Until then, check out our previous 2023 Progress Update event at livestream.newlimit.com. That link will also host our 2024 Update as the date gets closer.

Watch our livestream

Hao’s work was one of the inspirations for our program. Over the past five years, his group has developed a set of genomic tools to explore the effect diverse genetic perturbations on hepatocyte function in vivo. Rather than asking one question at a time, his team demonstrated that hundreds to thousands of hypotheses can be tested in a single mouse liver using pooled screening approaches. These experiments have already revealed disease mechanisms and therapeutic targets.

We’re excited to work with Hao to develop reprogramming medicines for the aging liver using similar approaches to explore many reprogramming interventions in parallel.

About Hao Zhu, MD

See Hao’s profile from the University of Texas Southwestern, paraphrased below

Hao is a Professor at the University of Texas Southwestern where he holds the The Nancy B. and Jake L. Hamon Distinguished Chair in Therapeutic Oncology Research. In addition to running a prolific research laboratory, Hao cares for hepatocellular carcinoma patients at Parkland Memorial Hospital.

Using the liver as a model system, the Zhu laboratory is trying to elucidate how injury, regeneration, and wound healing influence organ function and cancer formation. They believe that understanding somatic mosaicism will be a key genetic strategy to deconvolute the complexity of wound healing and are using human genomic approaches, in vivo genetic screening, and lineage tracing to understand the functional implications of somatic mosaicism in chronic liver disease and liver cancer.

Prior to UT Southwester, Hao earned his bachelor’s degree in biology from Duke University, followed by an M.D. from Harvard Medical School and MIT. He underwent clinical training in internal medicine at the University of California, San Francisco, and medical oncology at the Dana-Farber Cancer Institute. Dr. Zhu completed postdoctoral research in George Q. Daley’s laboratory at Boston Children’s Hospital.

We’ve spent the past two months extracting insights from these data and preparing follow up experiments to test new reprogramming factor sets based on the results. In parallel, we’ve begun laying the groundwork for our Metabolism program by onboarding hepatocytes to the Discovery Engine.

A few numerical highlights:

• >10X more reprogramming factor sets tested than our prior largest experiment

• >2M cells profiled, 10TB+ of raw sequencing data generated

• +3 team members

• 2X reduction in the time from in silico reprogramming predictions to experiment ready libraries

• +1 cell types demonstrated for our cell genomics chemistry & cell age prediction models

• 1st rejuvenating sets nominated

Subscribe to the NewLimit blog!

People

We welcomed several talented new members to the team at the start of the year:

• Nick Bernstein joined as a Senior Data Scientist on our Predictive Modeling team

• Justin Juang joined as a Research Associate on our Immunology team

• Megan McMaster joined as a Laboratory Operations Manager

Our Discovery Engine delivered the largest reprogramming screen to date

By the numbers, we captured more than 2 million cell profiles treated with thousands of partial reprogramming factor sets. This represents >50X more reprogramming sets tested than field prior to NewLimit’s founding and >10X more than in our previous largest experiment. By cell count, this single experiment is in the same range as the 10 largest datasets in the CZI Census compendium of single cell genomics data.

This experiment represents the first full cycle of our Discovery Engine. We trained models on prior screens, predicted the results of reprogramming interventions, and used those predictions to design our largest experiment yet. We believe this synergy of atoms & bits will enable us to search through the massive space of reprogramming factor sets to discover new medicines.

Large datasets emerging from our Engine are only valuable if they lead to biological insights. We were heartened to find that this latest screen revealed new sets of factors that reprogrammed cell age and function. We also found that our Engine reproduced prior reports of reprogramming effects on cell fitness, confirming the technical validity of our approach. We’re excited to evaluate these new factor sets in the coming months to see which of them make old primary cells act young.

Reading epigenetic states in a new cell type

We built our Discovery Engine to be largely cell type agnostic. Our reprogramming factor delivery tools, single cell profiling chemistries, and predictive models are not tied to any one lineage. Nonetheless, we had not challenged our Engine to adapt to a new cell type until January this year.

We launched our Metabolism program at the end of 2023 and began work to onboard hepatocytes to the Engine. In early experiments this January, our Single Cell Genomics team demonstrated that our existing cell profiling chemistries were capable of generating high quality data in hepatocytes, on par with the best published data in the academic literature. Even with a modest data corpus from our first experiment, our Predictive Modeling team found that we can reliably predict cell age from hepatocyte profiles.

These experiments are the first time we actually demonstrated that components of our Engine are cell type agnostic. Here alone, we showed that both our genomics chemistry and predictive models are transferable. By the end of the quarter, we’re optimistic that we’ll demonstrate the remaining components in hepatocytes as well.

Ensuring reproducibility

Functional genomics experiments like we run with our Discovery Engine can suffer from reproducibility challenges. From day one, we’ve been focused on ensuring that our results can be replicated time and again through the use of rigorous operating procedures, standardized parts, and a set of internal controls.

We recently had the opportunity to put the reproducibility of our Engine to the test when our Epigenetic Editing team found a way to shorten the timeline of our screens by 40%. We’re always excited by process improvements that can help us accelerate discovery, but it wasn’t immediately clear if the new approach would be as robust as our previous system.

We were pleased to find that the process we’ve built ensured that not only was each timeline reproducible across independent experiments, but the effects we observed were consistent even across these distinct experimental conditions.

Results like this give us confidence that our Engine is tuned and reliable.

Join us

We continue to recruit talented scientists, engineers, and operators. Please reach out if you’re excited by our mission to develop reprogramming medicines for aging!

A few highlights from the year:

• 50X more reprogramming factor sets tested in partial reprogramming experiments than the field has explored cumulatively

• 20X increase in the signal-to-noise of our pooled screen demultiplexing assay

• 7X the number of factors we can deliver to human cells

• 2X the number of NewLimit team members (12 -> 24)

• +3 functional assays to discriminate human T cell age

• +2 metrics where our in silico reprogramming models surpassed state-of-the-art

Subscribe to the NewLimit blog

Writing epigenetic states

In order to discover partial reprogramming factors, we first have to deliver these factors to old, human cells alongside the command-and-control machinery that allows us to turn them on and off. Our team spent the year’s opening months to make this aspiration a reality.

Delivering transgenes to real cells from human donors is a known hard problem that has plagued scientists and drug developers for decades. Most large scale genetic screens have avoided solving it altogether by restricting their experiments to cancerous cell lines that are eager to take up foreign genetic material. Our work requires more than just a passable solution to this problem. We need to get combinations of reprogramming factors into old cells, not just one at a time, and we need tight control over expression of those factors.

Our team was undeterred by this challenging history and iterated across dozens of delivery chemistries on our way to a solution. We turned our main conference room into a war room and treated each new piece of data like a dispatch from the frontier. After months of concerted effort, we had built tools that achieved all the metrics we needed for high performance discovery screens and more. We also uncovered previously unknown “rules” of the molecular systems we employ that have enabled new designs in the months since.

Reading designs from nucleotides

The classic way to test many partial reprogramming factors would be to put each unique combination into old cells in separate test tubes, then measure what happens. This encodes the design of the experiment in physical space. This would limit the number of factor sets we could test to the number of test tubes we could line up and manipulate in the lab.

NewLimit’s Discovery Engine instead delivers reprogramming factors to a pool of old cells all at the same time. The stochastic nature of life ensures that each cell randomly receives a different combination of factors. In order to figure out which factors each cell receives, we pair them with nucleic acid barcodes that we can measure later. Rather than encoding the design in physical space, we encode it in nucleotides.

Analyzing these complex experiments requires us to read these barcodes with high fidelity alongside a measurement of the reprogramming effects that we capture with single cell genomics. Any barcode that isn’t detected properly represents an injection of noise and inefficiency into our experiments. 

At the beginning of the year, the best methods available for our system detected somewhere between half and three-quarters of the barcodes present in each cell. Our team combed the scientific literature for tricks to improve performance – step by step, stacking one trick on top of another, a series of small improvements compounded to achieve a detection sensitivity >95% and specificity >99%. The overall signal to noise ratio of this read-out improved >20X from January to December. 

Our team further took this chemistry and developed a new sequencing technology we call pxbc-seq. To our knowledge, this technology is capable of measuring these “perturbation (px) barcodes (bc)” at single cell resolution with higher fidelity and lower costs than ever before. This technology unlocks larger, selection-based screening experiments that will allow us to test 10^4 reprogramming sets in a single dish.

We believe these chemistry improvements uniquely enable us to test more combinations of reprogramming factors than has been possible previously.

Revving the Discovery Engine

Together, these molecular tools allowed us to execute the largest partial reprogramming experiments in the world, to our knowledge. In July, we executed the first experiment to test more than 100 combinations of partial reprogramming factors. In October, we completed the first experiment that tested more than 100 unique factors. In December, we completed the first screen that tested more than 1000 factor combinations.

To our knowledge, the partial reprogramming field had tested <20 combinations of factors prior to NewLimit’s founding. The largest single experiment tested 16 sets. We can now test nearly two orders-of-magnitude more sets in a single experiment than was possible at the beginning of 2023. 

Each of these discovery screens has allowed us to accumulate a large data corpus of partial reprogramming effects to enable our predictive modeling efforts.

Predicting reprogramming outcomes in silico

No matter how sophisticated our experimental tools, we will never be able to test all of the reprogramming factor sets that are possible. There are simply far too many combinations to explore exclusively in the world of atoms. To help us search this massive space of possibilities, NewLimit is developing in silico reprogramming models to predict the effect of partial reprogramming interventions in the world of bits.

Our growing data corpus enabled us to build better models than ever before in 2023. Our in silico reprogramming model takes as input a set of reprogramming factors and a starting cell state, then predicts the phenotype that results from partial reprogramming. Our models now explain a great deal of the variation we’ve observed and, to the best of our knowledge, achieve state-of-the-art performance.

We’re now guiding our future experiments using these models alongside the judgment of our scientists.

Validating therapeutic utility

Our Discovery Engine reveals partial reprogramming factors that make old cells look like young cells based on their molecular phenotype. Developing therapeutics requires us to make old cells act like young cells, performing their obligate functions more effectively.

Our Immunology team has been heads-down throughout the year implementing a suite of functional assays that can distinguish functional differences between young and aged human T cells, then determine if those functions are improved by reprogramming. In January, we had 0 functional assays running and no real ability to measure the impact of our reprogramming interventions beyond apparent phenotypes. 

By December, our team had implemented a half-dozen assays for T cell function that measure everything from a cell’s ability to sound the alarm on an infection, to the potential of each cell to coordinate a humoral response. By combining a few of the assays, we have begun to tell old and young T cells apart based on their functions, as well as their phenotypes. 

As our Discovery Engine begins to emit promising interventions, our team is now prepared to test them with rigor.

2024

We’ve built the Discovery Engine we first envisioned at NewLimit’s inception. We’re now in a position to run the Engine and make discoveries to drive our therapeutic programs. Each month, we grow increasingly confident that age is reversible.

We’re merely at the beginning of a long road to new medicines, but we take pride in how rapidly we’re accelerating. Come join us!

These preparations have yielded a series of improvements to our Discovery Engine, which we are continually looking to scale. Each of these iterations compounds with those before it, so that even small wins, month over month, have improved how quickly we can search the space of new sets of reprogramming factors. We look forward to sharing highlights from our larger screens later this year.

Subscribe to the NewLimit blog

People

We’ve had the privilege of welcoming four new team members to NewLimit over the past two months:

• Sophie Liu joined as a Research Associate on the Write team.

• Neal Ravindra joined as a Machine Learning Scientist on the Predict team.

• Jordan Spice joined as a Research Associate on the Write team.

• Sarah You joined as an Operations Assistant on the Operations team.

Atoms

Expanding the palette of partial reprogramming factors

Near the end of the summer, we began constructing a partial reprogramming factor library that was 6X larger than any we had built to date. This month, we completed that construction process and successfully completed our first discovery screen using these new factors.

To the best of our knowledge, this is the largest partial reprogramming library that has been screened to date, and this experiment alone more than 3Xs the number of partial reprogramming factors that have been tested for their impact on cell age.

Raising signal above the noise

Our Discovery Engine delivers a pool of reprogramming factors to a pool of cells. Based on the stochasticity of life, each cell receives a different combination of factors from the pool. We then profile cells with single cell genomics so that each cell provides it’s own independent datapoint. This approach allows us to screen many possible combinations in a single culture dish.

One challenge with this approach is that we need to determine which factors are in which cells post-hoc at high fidelity. It turns out this is a rather hard problem! We’ve built a system of DNA barcodes and a companion single cell genomics assay to perform this demultiplexing step during our screening process.

This month, we were able to make a dramatic improvement in the signal-to-noise ratio of that demultiplexing assay. In practice, this means that we can perform screens with more complex combinations of reprogramming factors in each cell while maintaining high fidelity assignment of reprogramming factor combinations to cells.

Enabling ensemble screens for partial reprogramming factors

Most of our screening efforts to date have employed single cell profiling as the endpoint. For each set of reprogramming factors in each cell, we’ve measured a complex phenotype like the expression of every gene, or the accessibility of every region in the genome. These assays are powerful and we can scale them to thousands of factor sets per experiment.

In order to screen tens of thousands of factor sets at the same time though, we’re building an assay that makes different trade-offs. Rather than collecting a complex profile as the endpoint, we simply measure the abundance of each combination of factors in a pool of cells. If we pair this simple measurement with some way to select for cells that look or act more youthful after reprogramming, we can interrogate the effect of many combinations on this selectable feature at an even larger scale than our single cell screens.

This general approach is known as ensemble (“bulk”) screening, and it has proven incredibly valuable in the broader functional genomics community. However, a key challenge we face is that existing methods are unable to detect if reprogramming factors are acting in combination!

Given the importance of combinatorial logic for reprogramming, we’ve been developing a custom next generation sequencing assay to measure the abundance of reprogramming factor combinations in an ensemble format. Our first v0 approach slightly improved on our ability to detect combinations relative to existing methods, but sacrificed our ability to detect single factors in each cell accurately.

This month, we deployed our v1 approach that matches the single factor detection performance of a baseline and achieves much higher performance for detecting factor combinations. We’ll be putting this new assay into production in the coming months.

Increasing Discovery Engine RPM

Our Discovery Engine implements a design-build-test-learn cycle in which we test reprogramming factor combinations for their ability to reverse cell age, then learn from the results to suggest new factors to test.

One of our high level hypotheses is that repeating this cycle more times will increase our chances of success. This month, we focused on shortening the timeline required to execute one of our reprogramming discovery screens in order to fit more engine cycles into each unit of time. Our team managed to reduce the time per cycle by 40%, commensurately increasing our Engine RPM by 1.6X.

Bits

In silico reprogramming at higher fidelity

We’ve been developing in silico reprograming models to predict the effect of a reprogramming factor set on cell age since the beginning of 2023. Over the past two months, we’ve made meaningful improvements to these models that increase the correlation between observed and predicted reprogramming effects. At this stage, we believe our models can predict reprogramming effects as well or better than the prior state-of-the-art (SOTA).

Inferring cell age across diverse domains

Earlier this year, we developed machine learning models that learn to infer cell age — the age of the organism the cell came from — from cell profiles. We’ve previously observed strong performance across a range of conditions, but in the past two months, we spent more time than ever putting these models through their paces and evaluating robustness.

Biological data frequently suffers from domain shifts (aka “batch effects”) where the target data for predictions is a bit different than any of the training data. This can make it challenging to develop models that are robust against technical changes and consistently provide high performance. We put our models through a gamut of challenging domain shift tasks, including predicting on cells from different subtypes, predicting age across experimental conditions, and predicting on entirely distinct sequencing technologies.

We were happy to find that our models performed well, in spite of all the challenges we threw at them. Experiments like these are essential to build confidence in our in silico phenotyping tools, and we’ll continue to act as our own red team against our models in the days to come.