🧬

NewLimit’s Approach to Longevity

Created
May 13, 2025 2:57 PM
Topics
longevity
Author
vasa

This is a working draft.

We all sense the passage of time, don't we? But what exactly is aging? Is it simply the accumulation of years, or is there something more fundamental happening within our very cells?

Scientists have been grappling with this question for ages, leading to an array of ideas about how and why we age. You might have come across some of these approaches – perhaps strategies focused on diet, exercise, or even more radical interventions aimed at extending our lifespan (all the things Bryan Johnson does).

But what if we could tackle aging at a more fundamental level, right down to the cells themselves? Imagine our cells carrying not just the blueprint of our being (our DNA), but also a kind of cellular "age marker." What if we could learn to manipulate these markers? This is where the groundbreaking work of companies like NewLimit comes into play. They're taking a unique "reprogram approach," diving deep into the realm of epigenetics to understand and potentially reverse cellular aging. How exactly do they aim to achieve this? Let's delve into their fascinating strategy.

Is there any indication that aging can be reversed on a cellular level?

image

Here's how we can connect the diagram to SCNT:

  • Embryonic Cell: In SCNT, the process starts with an oocyte (an egg cell). The nucleus (containing the DNA) of this oocyte is removed. So, the "Embryonic Cell" in the diagram represents this enucleated oocyte.
  • Aging Cell: The "Aging" cell in the diagram refers to a somatic cell (any cell of the body other than sperm or egg cells) taken from an organism. In the PDF, they used cumulus cells (cells surrounding the egg in the ovary).
  • Reprogramming (SCNT): The crucial step of SCNT involves taking the nucleus from the "Aging" somatic cell and transferring it into the enucleated oocyte. This is the "Reprogramming" stage in the diagram. The factors in the oocyte then act to "reprogram" the somatic cell nucleus, effectively rewinding it to an earlier developmental state.
  • Development: The oocyte, now containing the donor nucleus, is stimulated to start dividing and develop into an embryo. This develops into a new organism.

In essence, the diagram simplifies the complex process of SCNT to illustrate the concept of reversing the "age" of a cell by transferring its nucleus into an oocyte. The PDF then provides the scientific evidence and details of how this "reprogramming" is achieved through SCNT, specifically in mice.

Epigenetics

So, we were just talking about how NewLimit is looking at our cells and their "age tags." But what exactly are these tags they're so interested in? Have you ever thought about how it's possible for you to have so many different kinds of cells – skin cells, brain cells, liver cells – when they all contain the exact same DNA, the same fundamental blueprint? It's kind of mind-blowing when you think about it, right?

Well, this is where something called epigenetics comes into the picture. Think of our DNA as a massive instruction manual. Every cell in your body has the same manual. But a skin cell doesn't need to read the instructions for being a liver cell, and vice versa. So, how do they know which parts of the manual to read and which parts to ignore?

That's where these "epigenetic tags" come in. Imagine little sticky notes or labels that sit on top of the DNA.

The image shows a diagram of DNA with labels "Epigenetics" and shows tags on genes. (
The image shows a diagram of DNA with labels "Epigenetics" and shows tags on genes. (source)

These modifications don't change the underlying DNA sequence itself, but they act like switches, telling a gene whether to be "on" or "off," or somewhere in between.

So, now that we have a basic idea of what epigenetics is – these "sticky notes" on our DNA that control gene activity – we can start to think about how this relates to aging.

Scientists have already figured out how to manipulate cell age and cell type, as illustrated in the below diagram. By adding a specific set of four proteins called transcription factors (OCT4, SOX2, KLF4, and MYC) to an old skin cell, they can essentially hit the "reset" button and turn it into a young embryonic stem cell. Think of it as not just making the cell younger, but also giving it the potential to become any cell type in the body.

image

The next diagram shows another interesting possibility: we can also change the type of a cell – say, turning an old skin cell into a neuron or a liver cell – by using a different cocktail of transcription factors (1-6). In this case, the cell's identity changes, but it remains old.

image

Now, here's where NewLimit's approach, and the concept of partial reprogramming, becomes really exciting, as shown in the below diagram.

image

Their goal isn't to completely rewind a cell all the way back to a stem cell. Instead, they're aiming to discover how to change the age of a cell – to make an old skin cell a young skin cellwithout changing its fundamental identity. They want to find the right "recipe" of transcription factors (those "? TFs" in the diagram) that can rejuvenate the cell while it remains a skin cell. This "partial reprogramming" is like carefully turning back the clock on a cell without erasing its memory of what it's supposed to be.

Why is this partial approach so interesting? Well, completely reprogramming a cell back to a stem cell, while it makes it young, also erases its specialized function. A skin stem cell isn't doing the same job as a mature skin cell that protects our body. Partial reprogramming offers the potential to rejuvenate cells while keeping their essential functions intact.

So, the big question then becomes: how do you actually find these magic transcription factors that can partially reprogram a cell? That's where NewLimit's innovative "discovery engine" comes into play, which we can explore next.

Okay, so we've established that NewLimit is on the hunt for these "magic" transcription factors to partially reprogram cells. But how do they actually find them? That's where their clever "discovery engine" comes into play.

image

Think of it as a multi-stage search, where they start with a huge number of possibilities and gradually narrow it down. The image you provided outlines this process:

  1. In silico Screening: They begin with a massive number of hypotheses (permutations of TFs) – around 101010^{10} . This is like casting a very wide net, using computer simulations to predict which transcription factors might have the desired effect.
  2. Ensemble Pooled Screening: Next, they move to ensemble pooled screening, reducing the number of hypotheses to 10510^5. This involves testing groups of transcription factors together to see if they have any effect on cell age.
  3. Single Cell Pooled Screening: They then refine the search further with single-cell pooled screening, bringing the number of hypotheses down to 10310^3. Here, they look at the effects of transcription factors on individual cells, giving them more precise data.
  4. Cell type-specific Functional Assays: After this, they move to cell type-specific functional assays, reducing the number of hypotheses to 10210^2. This step involves testing the most promising transcription factors in specific cell types (like skin cells or liver cells) to see if they truly rejuvenate them while maintaining their function.
  5. Indication-specific Preclinical Models: Finally, they test the most promising candidates in preclinical models, further narrowing the field to 10110^1. This is where they start to see if these factors can have an effect in more complex systems.

It's a bit like a funnel, starting with a huge number of ideas and gradually filtering them down to the most promising candidates. This multi-step approach allows NewLimit to systematically explore the vast landscape of transcription factors and identify the "recipes" that can effectively and safely partially reprogram cells.

So, NewLimit has this sophisticated engine to find the right combination of factors. Now, let's think about how they actually test these factors. The next image, "How can we discover reprogramming factors?", gives us a glimpse into their innovative approach compared to more traditional methods.

image

The traditional way of searching for these reprogramming factors is quite laborious. It involves testing thousands of individual combinations in separate test tubes. Imagine having one test tube for factor A, another for factor B, then one for A and B together, and so on. As you can see, with potentially millions of combinations of transcription factors, this approach becomes incredibly time-consuming and expensive. The image points out that the costs scale linearly with the number of combinations. Also, given that aging is most likely a function of multiple genes, you need to measure the activity (read-out) of multiple genes (instead of a single one).

NewLimit is taking a much more streamlined approach. They're using a technique that allows them to test many combinations of TFs simultaneously in a single culture dish. They create a "reprogramming factor pool" – a mix of different potential factors. Then, they can introduce this pool into cells in a single dish and use advanced sequencing techniques to read out the effects of many different combinations at once. The image indicates they can get around test gene activity of multiple genes for each combination of TFs from a single dish, allowing them to test thousands of hypotheses in parallel.

This high-throughput screening dramatically speeds up the discovery process. Instead of laboriously testing each combination one by one, they can analyze the effects of many different "recipes" of transcription factors simultaneously. This allows them to quickly identify promising candidates that can then be further investigated in the later stages of their discovery engine.

So, after NewLimit uses their high-throughput screening to identify promising combinations of transcription factors, the next crucial step is to figure out which of these combinations are actually working to rejuvenate the cells. The image you've now shared, "How NewLimit's partial reprogramming screens work," gives us a glimpse into this process.

image

They start with human T cells, which are immune cells. They have both young and old T cells in their experiments. Then, they introduce their "TF Pool" – that mix of potential reprogramming factors we talked about. To control when these factors are active, they use a "Transgenic drug inducible DNA barcoded" system. This allows them to turn the factors on and off at specific times, giving the cells a "Reprog Pulse" – a period where the reprogramming factors are active – followed by a "Reprog Chase," where they observe the effects.

The key here is the "Single cell multi-omics" analysis. This advanced technique allows them to look at many different aspects of individual cells after they've been exposed to the reprogramming factors. They can measure things like gene expression (which genes are turned on or off), protein levels, and even the overall state of the cell.

Finally, they use "In Silico demultiplex & phenotype" to analyze this massive amount of single-cell data. By using computational tools, they can figure out which combinations of transcription factors (remember those DNA barcodes?) led to the most significant changes in the cells, ideally making the old cells look and act more like young cells. This step helps them identify the specific "recipes" that are truly effective in partially reprogramming the cells.

It's a sophisticated process of controlled activation of potential factors, followed by detailed analysis at the single-cell level to pinpoint the winners. This allows them to move beyond just seeing if something happens to understanding how and why it happens at a very granular level.

Alright, so after NewLimit has identified transcription factor combinations that seem promising in making old cells look younger based on their multi-omics analysis, the next crucial question is: do these cells also act like young cells? As the next image aptly puts it, "Function is the final boss."

Just looking younger isn't enough; the cells need to regain the functional capabilities of youthful cells. This is where NewLimit puts these partially reprogrammed cells through a series of rigorous functional assays. The image outlines several examples of these tests:

image
  • Proliferation (Acute stim.): Can the rejuvenated cells still divide and multiply like young cells when stimulated? This is a fundamental characteristic of youthful cells.
  • Persistence (Chronic stim.): Can the cells maintain their function and survival over longer periods, similar to young cells facing chronic stimulation?
  • Targeted Killing: In the case of immune cells like T cells (which they used in their initial screens), can the reprogrammed cells effectively target and eliminate threats, a key function of young, healthy immune cells?
  • Antibody Response: For other immune cells like B cells, can they still produce antibodies effectively when challenged, a crucial aspect of a functional immune system?
  • Stemness (Sorting): Do any of the reprogramming approaches induce a return to a more stem-like state within the specific cell type, which might be beneficial for regeneration?

These are just examples, and the specific functional assays NewLimit uses will depend on the type of cell they are trying to rejuvenate. The key takeaway is that they don't just rely on molecular markers that indicate a younger state; they rigorously test whether the cells can actually perform the jobs that young, healthy cells do.

This focus on function is critical because the ultimate goal isn't just to make cells look younger on a lab test, but to restore their youthful activity within the body, potentially leading to real benefits in terms of health and longevity.

Okay, so NewLimit has a sophisticated system for screening and functionally validating potential reprogramming factors. But where do they even begin to look for these factors in the first place? With potentially thousands of transcription factors in the human genome, how do they narrow down their search to a manageable number of promising candidates? The next image, "Where do we get reprogramming hypotheses?", sheds light on this crucial initial step.

image

The image highlights that NewLimit draws their initial hypotheses from several key sources:

  • Epigenetic Aging Maps: As we discussed earlier, the epigenetic landscape of our cells changes as we age. Scientists have been creating detailed maps of these changes in different cell types. NewLimit analyzes these "epigenetic aging maps" to identify specific epigenetic modifications that are consistently associated with aging. They then hypothesize that transcription factors which can reverse these age-related epigenetic changes might be good candidates for partial reprogramming. The image shows multi-omic profiles of T human T cells, illustrating different states like "age-exhausted" and "naive," suggesting they look for epigenetic signatures that distinguish these states.
  • TF Hypotheses: The image also shows a bar graph representing the number of cells sequenced over time, leading to a set of "TF hypotheses" (represented by circles A, B, C, and D). This implies that by deeply analyzing the molecular changes in cells (likely including gene expression and epigenetic modifications) across different ages and conditions, they can generate hypotheses about which transcription factors might be involved in regulating these changes. The arrow suggests a link between the large-scale sequencing data and the generation of specific TF hypotheses.

In essence, NewLimit uses the knowledge of how our epigenome changes with age as a compass to guide their search for reprogramming factors. By understanding the epigenetic signatures of aging, they can make more informed guesses about which transcription factors might have the power to reverse these changes and rejuvenate cells. This targeted approach significantly increases their chances of finding effective partial reprogramming strategies compared to a purely random search.

So, NewLimit has these promising leads on potential reprogramming factors. The next logical step is to scale up their testing to efficiently evaluate a larger number of these candidates. The image titled "NewLimit's throughput (2023)" illustrates their efforts in this direction.

image

The bar graph on the left, "Scaled screens @NewLimit," shows a significant increase in the number of reprogramming sets they've tested over time, particularly in early 2023. This indicates that they've developed the capacity to run their screening process on a much larger scale compared to their earlier efforts ("'16-'22 ex-NL"). This increased throughput is crucial for efficiently sifting through the many hypotheses generated from their epigenetic aging maps and initial screens.

The scatter plot on the right, "Partial reprogramming phenotypes," gives us a glimpse into the results of these scaled screens. Each dot on this plot likely represents a different combination of reprogramming factors they've tested. The position of a dot in this "embedding of reprogramming effects learned in a generative model latent space" suggests the overall impact of that factor combination on the cells' phenotype – their observable characteristics.

You can see clusters of dots, including positive and negative controls, as well as several "Novel" factor combinations. This visualization allows NewLimit's scientists to quickly identify which factor combinations are having the most significant effects on the cells, pushing them towards a more "positive control" phenotype (presumably resembling younger cells) and away from the "negative control."

This ability to perform scaled screens and visualize the resulting phenotypic changes in a meaningful way is a powerful tool in NewLimit's discovery engine. It allows them to rapidly iterate, learn from their experiments, and focus their efforts on the most promising avenues for partial reprogramming.

Okay, having established their ability to control the timing of reprogramming factors, the next step for NewLimit is to apply this knowledge in sophisticated screening assays. The image, "NLMT-seq: Pooled screens with temporal control of reprogramming factors," showcases one such advanced technique.

image

The diagram illustrates how they start with T cells and introduce a pool of inducible and DNA-barcoded transcription factors. By adding an inducer, they can turn on these factors. After a specific period, they can remove the inducer, effectively turning the factors off. At various time points during this "on-off" cycle, they perform single-cell genomics to see how the cells are responding at the molecular level. The DNA barcodes allow them to link specific transcription factor combinations to the observed changes in individual cells.

The box plots in the image likely represent the expression levels of the introduced transcription factors ("All TFs On" vs. "TFs off") and the detection of the barcodes ("TF Barcode Detected"). This confirms that they can effectively control the expression of their factor combinations and track their presence within individual cells.

This sophisticated approach allows NewLimit to answer critical questions like:

  • How quickly do different factor combinations induce changes in the cells?
  • What is the optimal duration of exposure to the reprogramming factors?
  • Are the observed changes stable after the factors are turned off?
  • Do different cell types respond differently to the same temporal patterns of reprogramming factors?

By meticulously studying the temporal dynamics of partial reprogramming at single-cell resolution, NewLimit aims to identify the most effective and safe strategies for cellular rejuvenation. This deep understanding of the process is crucial for translating their findings into potential therapies.

Building on their ability to control the timing of reprogramming factors and analyze the cellular responses in detail, NewLimit takes a crucial step: validating whether their reprogramming screens can actually recover known biological principles. The image, "Reprogramming screens recover known biology," illustrates this important validation step.

image

The diagram on the left shows a simplified representation of their temporal control experiment. When the transcription factors (TFs) are turned "on" (indicated by the lightbulb), and the barcode (BC) for those TFs is present, they observe the expression of target genes (indicated by the wavy lines). When the TFs are turned "off," the target gene expression diminishes.

The bar graphs on the right provide quantitative data supporting this observation. They show the expression levels of "Target 1" and "Target 2" under different conditions. Notice that the expression of these target genes is significantly higher when the transcription factors are active ("TFs Act.") compared to the negative controls ("Neg. Ctrls."). This demonstrates that their screening system can indeed detect and measure the expected biological effects of transcription factor activity.

According to the video, this step of "recovering known biology" is crucial for building confidence in their screening platform[cite: 9]. It's like calibrating a scientific instrument to ensure it's giving accurate readings. By showing that their system can reproduce well-established relationships between transcription factors and their target genes, NewLimit gains greater assurance that the novel results they obtain from their screens are likely to be meaningful and reliable.

This validation step strengthens the foundation of their research, allowing them to more confidently explore novel combinations of reprogramming factors and interpret the resulting changes in cellular phenotypes. It confirms that their sophisticated screening and analysis pipeline is indeed capable of uncovering meaningful biological insights relevant to cellular rejuvenation.

With their screening platform validated, NewLimit can now delve deeper into exploring the diverse effects of different partial reprogramming strategies. The image, "Partial reprogramming elicits diverse phenotypes," showcases some of the varied outcomes they observe.

image

Each of the six plots in the image represents a different condition or set of reprogramming factors they've tested. These are likely visualizations of high-dimensional single-cell data, similar to the "Partial reprogramming phenotypes" scatter plot we saw earlier, but perhaps focusing on different cellular features or using a different dimensionality reduction technique. The contours within each plot represent clusters of cells with similar characteristics.

Notice the diversity in these plots. The "Negative Control" shows one distribution of cellular states, while the "Positive Control" (likely representing cells treated with known rejuvenation factors or young cells) shows a different distribution. The "Yamanaka Factor" plot, representing the full Yamanaka factors used for complete reprogramming, shows yet another distinct pattern, potentially highlighting the shift towards a more stem-like state.

Crucially, the "Novel 1," "Novel 2," and "Novel 3" plots each show unique distributions of cellular states. This suggests that different combinations of the novel transcription factors NewLimit is testing are eliciting a variety of cellular responses. Some might be nudging the cells closer to the "Positive Control" phenotype, indicating successful rejuvenation, while others might be having different or less pronounced effects.

As the video likely explains, this diversity of phenotypes is expected and is a key part of the discovery process[cite: 10]. By systematically exploring a wide range of factor combinations and carefully analyzing the resulting cellular states, NewLimit can identify the specific "recipes" that most effectively and safely induce partial reprogramming, moving cells towards a youthful phenotype without causing unwanted changes in cell identity.

This detailed phenotypic analysis allows them to move beyond simply identifying whether a factor has an effect to understanding how it's affecting the cells and selecting the most promising candidates for further development.

To ensure a coherent flow and introduce new information, let's revisit the concept of diverse phenotypes and connect it more explicitly to NewLimit's overall strategy as potentially outlined in the video.

As we've seen, NewLimit's screens reveal a range of cellular responses to different reprogramming factor combinations. This diversity is not seen as a setback but rather as a rich source of information. The video likely emphasizes that by meticulously mapping these diverse phenotypic outcomes, NewLimit can gain a deeper understanding of the underlying biology of cellular aging and rejuvenation[cite: 10].

image

Each unique cluster of cells in the "Partial reprogramming elicits diverse phenotypes" plot potentially represents a distinct state of partial reprogramming. By characterizing the molecular features of cells within each cluster, NewLimit can identify the specific changes induced by different factor combinations. This allows them to:

  • Identify optimal rejuvenation strategies: By comparing the phenotypes induced by novel factors to those of young cells (the positive control), they can pinpoint the combinations that most effectively reverse the aging process.
  • Understand the mechanisms of action: Analyzing the molecular changes associated with different phenotypes can reveal the specific pathways and genes involved in cellular rejuvenation.
  • Minimize off-target effects: By carefully examining the full spectrum of phenotypic changes, they can identify factor combinations that primarily induce rejuvenation without causing undesirable alterations in cell identity or function.
  • Develop targeted therapies: The insights gained from studying diverse phenotypes can inform the development of highly specific and effective partial reprogramming therapies for different cell types and age-related conditions.

In essence, the exploration of diverse phenotypes is a cornerstone of NewLimit's approach. It allows them to move beyond a "one-size-fits-all" mentality and develop a nuanced understanding of how to precisely manipulate cellular age for therapeutic benefit.

image

Let's elaborate on how machine learning enhances NewLimit's experimental process, drawing from the video.

image

The image "Machine learning extracts more value from each experiment" visually summarizes how NewLimit uses machine learning to analyze their complex data. The image shows how machine learning is used to extract more value from each experiment. It depicts the following:

  • Multi-perturbation barcode calling: The image shows how different transcription factors (represented by colored circles) are used to perturb cells, and how barcodes are used to track these perturbations.
  • Data-driven cell typing: Machine learning is used to classify cells into different types or states (aged vs. young) based on their molecular profiles. The image shows a graph depicting the transition from aged to young cells.
  • In silico perturbation screening: Machine learning is used to predict the effects of different transcription factor combinations. The image shows how different transcription factor combinations are fed into a machine learning model to predict their effects.

The video likely explains that NewLimit's experiments generate vast amounts of single-cell data, far too complex for traditional analysis methods. Machine learning algorithms are crucial for extracting meaningful insights from this data. By training models on their experimental results, NewLimit can:

  • Identify key features: Machine learning helps pinpoint the specific molecular changes (gene expression, epigenetic modifications, etc.) that are most strongly associated with successful cellular reprogramming.
  • Predict effective factor combinations: The models can predict which combinations of transcription factors are most likely to induce the desired rejuvenation effects, allowing NewLimit to prioritize their experimental efforts.
  • Improve experimental design: Machine learning can suggest optimal experimental parameters, such as which factors to test, at what concentrations, and for how long, to maximize the information gained from each experiment.

In essence, machine learning acts as a powerful lens, allowing NewLimit to extract significantly more value from each experiment and accelerate their search for effective partial reprogramming strategies.

Let's elaborate on how machine learning enhances NewLimit's experimental process, drawing from the video.

image

The image "Machine learning extracts more value from each experiment" visually summarizes how NewLimit uses machine learning to analyze their complex data. It depicts the following:

  • Multi-perturbation barcode calling: The image shows how different transcription factors (represented by colored hexagons) are used to perturb cells, and how barcodes are used to track these perturbations.
  • Data-driven cell typing: Machine learning is used to classify cells into different types or states (aged vs. young) based on their molecular profiles. The image shows a graph depicting the transition from aged to young cells.
  • In silico perturbation screening: Machine learning is used to predict the effects of different transcription factor combinations. The image shows how different transcription factor combinations are fed into a machine learning model to predict their effects.

The video likely explains that NewLimit's experiments generate vast amounts of single-cell data, far too complex for traditional analysis methods. Machine learning algorithms are crucial for extracting meaningful insights from this data. By training models on their experimental results, NewLimit can:

  • Identify key features: Machine learning helps pinpoint the specific molecular changes (gene expression, epigenetic modifications, etc.) that are most strongly associated with successful cellular reprogramming.
  • Predict effective factor combinations: The models can predict which combinations of transcription factors are most likely to induce the desired rejuvenation effects, allowing NewLimit to prioritize their experimental efforts.
  • Improve experimental design: Machine learning can suggest optimal experimental parameters, such as which factors to test, at what concentrations, and for how long, to maximize the information gained from each experiment.

In essence, machine learning acts as a powerful lens, allowing NewLimit to extract significantly more value from each experiment and accelerate their search for effective partial reprogramming strategies.

Let's elaborate on how machine learning enhances NewLimit's experimental process, drawing from the video.

image

As we've discussed, NewLimit generates a lot of data from their experiments. To efficiently analyze this data and refine their search for reprogramming factors, they employ a "Design-Build-Test-Learn" cycle. This iterative process, guided by machine learning, allows them to rapidly improve their reprogramming strategies.

The image "Rapid Iteration via Design-Build-Test-Learn Cycles" depicts several parallel screening experiments, each represented by a circle. The colored portions of the circles indicate the relative time spent on different phases of the cycle:

  • Design: Planning and designing the experiment.
  • Build: Constructing the experimental setup and generating the reprogramming factors.
  • Test: Running the experiment and collecting data.
  • Learn: Analyzing the data and using machine learning to inform the next design.

The image shows examples of different types of screens, including:

  • High-throughput single-cell CITE-seq test
  • Pooled mRNA screen in Jurkats (a type of T cell)
  • Pooled mRNA screen in primary T cells
  • Matched arrayed and pooled mRNA screen in Jurkats
  • Pooled mRNA screen in human T cells

The video likely emphasizes how machine learning accelerates each stage of this cycle. For example, machine learning can help:

  • Design new experiments by predicting which transcription factor combinations are most likely to be effective.
  • Analyze the complex data generated by their single-cell sequencing experiments.
  • Learn from previous experiments to refine their models and improve the design of future screens.

This rapid iteration, driven by machine learning, allows NewLimit to quickly explore the vast landscape of potential reprogramming factors and identify the most promising avenues for cellular rejuvenation.

Let's elaborate on how machine learning enhances NewLimit's experimental process, drawing from the video.

image

The image "Functional outcomes of TF reprogramming can be predicted in silico" illustrates how NewLimit uses machine learning to predict the functional outcomes of transcription factor reprogramming.

The image depicts the following:

  • TF pairs: Different pairs of transcription factors (represented by A+B, C+B) are used to perturb cells.
  • Function: The image shows the functional outcome of these perturbations (represented by upward or downward arrows).
  • Test-set performance: The graph shows how well a machine learning model can predict the functional outcome of new transcription factor perturbations.

The video likely explains that NewLimit uses machine learning to predict which transcription factor combinations will lead to desired functional outcomes, such as improved cell proliferation or targeted killing. By training models on their experimental data, they can identify patterns and relationships between transcription factor combinations and cellular function. The graph shows that their model outperforms random search, indicating that it can effectively predict functional outcomes. This "in silico" experimentation allows NewLimit to prioritize the most promising factor combinations for further testing, saving time and resources.

The graph, titled "Test-set performance: predicting positive function from new perturbations," is about evaluating how well NewLimit's machine learning model can predict whether a given combination of transcription factors will lead to a "positive" functional outcome in cells.

Here's a step-by-step explanation:

  • X-axis: Fraction of experiments explored: This axis represents how much of their experimental data NewLimit is using. 0.0 means they're using none, and 1.0 means they're using all of it. Think of it as how much "learning" the machine learning model has done.
  • Y-axis: Fraction of positive hits found: This axis shows the proportion of successful experiments (those with a "positive" functional outcome) that the model correctly identifies. A "positive hit" is an experiment where the transcription factor combination did achieve the desired functional outcome (e.g., increased cell proliferation, enhanced immune response).
  • "Our model (AUC = 0.65 ± 0.01)": The blue line represents the performance of NewLimit's machine learning model. The "AUC" stands for "Area Under the Curve," which is a common metric to evaluate the overall performance of a classification model. An AUC of 0.5 indicates that the model is no better than random chance, while an AUC of 1.0 indicates perfect prediction. In this case, an AUC of 0.65 suggests the model performs moderately well, better than chance but not perfect. The "± 0.01" indicates the standard error, showing the precision of the AUC measurement.
  • "Random search": The dashed black line represents what you'd expect if you were just randomly guessing which transcription factor combinations would work. This line goes diagonally from the bottom left to the top right. It means that if you randomly picked 20% of your experiments, you'd expect to find 20% of the positive hits, and so on.
  • Interpretation: The key takeaway is that the blue line (NewLimit's model) is above the dashed black line (random search). This means that their machine learning model is significantly better at finding "positive hits" than simply guessing. For example, if they explore only 20% of possible experiments (x-axis = 0.2), their model helps them find a larger fraction of the actual positive hits (y-axis is higher for the blue line than the black line).

In simpler terms, the graph shows that NewLimit's machine learning model is effective at predicting which transcription factor combinations are most likely to work, allowing them to focus their experimental efforts and find successful rejuvenation strategies more efficiently.

Let's elaborate on how machine learning enhances NewLimit's experimental process, drawing from the video.

image

The image "In silico search can prioritize next experiments" visually summarizes how NewLimit uses machine learning to guide their experimental design.

The image depicts the following:

  • Measure: Initial experiments are conducted to measure cellular responses to different transcription factor combinations.
  • Predict: Machine learning models are trained on this data to predict the effects of new transcription factor combinations.
  • Prioritize: The model's predictions are used to prioritize which experiments to conduct next, focusing on the most promising factor combinations.
  • In silico screening: The image shows a graph depicting the probability of a novel strategy improving function. The graph indicates that NewLimit performs in silico screening of 1.8M reprogramming factor sets.

The video likely explains that NewLimit uses machine learning to predict which transcription factor combinations will lead to desired functional outcomes, such as improved cell proliferation or targeted killing. By training models on their experimental data, they can identify patterns and relationships between transcription factor combinations and cellular function. This "in silico" experimentation allows NewLimit to prioritize the most promising factor combinations for further testing, saving time and resources. The image illustrates that the probability of a novel strategy improving function is much higher than random chance.

The graph in the "In silico search can prioritize next experiments" image is designed to illustrate the effectiveness of using machine learning to guide experimental design in the search for reprogramming factors.

Here's a more detailed explanation:

  • X-axis: Fraction of Experiments Explored: This axis represents the proportion of all possible experiments that NewLimit could perform. It ranges from 0 (meaning no experiments have been done) to 1 (meaning all possible experiments have been done). Think of it as how much of the "search space" they've covered.
  • Y-axis: Probability of a Novel Strategy Improving Function: This axis represents the likelihood that a new combination of reprogramming factors (a "novel strategy") will actually lead to an improvement in cellular function (e.g., increased proliferation, enhanced clearance of damaging proteins). It ranges from 0 (no chance of improvement) to 1 (certainty of improvement).
  • The Curve: The blue curve shows the probability of finding a successful strategy as NewLimit explores more of the possible experiments, guided by their machine learning model.
  • The Horizontal Line: The horizontal line represents the probability of success if NewLimit were to choose experiments randomly.

Key Interpretation:

The most important thing to notice is that the blue curve is significantly higher than the horizontal line. This demonstrates the power of machine learning in prioritizing experiments.

  • If NewLimit were to randomly select experiments, the probability of finding a successful strategy would remain relatively low and constant, as shown by the horizontal line.
  • However, by using machine learning to analyze the results of previous experiments and predict which new strategies are most promising, NewLimit can significantly increase their chances of success. The blue curve shows that, even early on, their model helps them identify more promising avenues of research.

In essence, this graph provides a visual representation of how machine learning enables NewLimit to navigate the vast space of possible reprogramming factor combinations much more efficiently than if they were relying on chance. It highlights the value of their computational approach in accelerating the discovery of effective anti-aging therapies.

Let's discuss NewLimit's vision for how their research might translate into actual therapies.

image

The image "What will our medicine look like?" suggests that NewLimit envisions using liquid nanoparticles (LNPs) with mRNAs that encode their reprogramming factors.

Here's why this is significant:

  • LNPs as delivery vehicles: LNPs are tiny spheres made of lipids (fats) that can encapsulate and protect fragile molecules like mRNA. They are already used in some approved medicines, including mRNA vaccines, and can deliver their cargo into cells.
  • mRNA as the instructions: mRNA (messenger RNA) carries the genetic instructions for making proteins. In this case, the mRNA would encode the reprogramming factors – those special proteins that can partially rejuvenate cells.
  • Scalable and low-cost manufacturing: The image highlights that LNP + mRNA drugs can be manufactured at scale with low cost. This is crucial for making potential therapies accessible to a large population.

The video likely explains that this approach allows for a relatively straightforward way to deliver the reprogramming factors into cells. Instead of directly introducing the proteins, which can be challenging, they can use mRNA to instruct the cells to produce the factors themselves. This method is also potentially safer and more controllable than other gene therapy approaches.

In essence, NewLimit is exploring a cutting-edge approach to deliver their reprogramming factors, leveraging the power of mRNA and the versatility of LNPs. This strategy could pave the way for future therapies that target aging at its source: the cellular level.

Let's explore the therapeutic potential of NewLimit's research, focusing on the possibility of treating disease by rejuvenating T cells.

image

The image "Would rejuvenating T cells treat disease?" suggests that rejuvenating T cells could be a promising therapeutic strategy.

Here's how the image and video likely connect:

  • Young vs. Aged T Cells: The image highlights the difference between young and aged T cells. Young T cells are more effective at fighting off viruses. The graph shows that young animals with aged T cells are susceptible to viruses, whereas young animals with young T cells are protected.
  • T cell responses to vaccines: Aged animals with young T cells respond better to vaccines. The graph shows that young T cells improve responses in aged animals.

The video likely discusses how NewLimit's partial reprogramming approach could be used to rejuvenate aged T cells, restoring their ability to fight infections and respond to vaccines. This could have significant implications for treating age-related immune decline and improving outcomes for elderly individuals.

Let's discuss NewLimit's vision for how their research might translate into actual therapies, specifically focusing on T cell reprogramming.

image

The image "Features of T cell reprogramming drug program" highlights the potential of T cell reprogramming to impact various age-related conditions, ranging from infectious diseases to cancer.

Here's how the image and video likely connect:

  • Robust in vitro functional assays and high-throughput systems: NewLimit uses sophisticated methods to test the effectiveness of their T cell reprogramming strategies.
  • Impact on diverse age-related indications: T cell rejuvenation holds promise for treating a wide range of age-related diseases, including infectious diseases and cancer.
  • Scalability: T cell applications align with NewLimit's ambitious goals.

The video likely explains that NewLimit is specifically interested in T cells because of their crucial role in the immune system. As we age, our T cells become less effective at fighting off infections and cancer. By rejuvenating these cells, NewLimit hopes to restore youthful immune function and improve healthspan.

image

A few delivery modalities:

At the end of the day, we need the host cells to have the transcription factors (proteins; in most cases) which will alter the epigenetic landscape of the host cell.

So, there are different ways to do it;

  • send the genes that encode the TFs using viral vectors. The most commonly used vectors are from lentiviruses and adeno-associated viruses.
  • you can also use mRNA (instead of the genes). LNPs for delivering mRNA.
  • there are mice lines bred which have the genes that encode the TFs already.
resources
Q/As
questions