November 2022 Progress Update
Arrival at the starting line
As we near the end of 2022, our whole team has been heads-down closing out our annual goals. We’re excited to share our progress here and even more excited to share the outputs in the new year.
Subscribe to our newsletter to stay updated on our progress in developing epigenetic reprogramming medicines.
Why share progress updates?
NewLimit is pursuing a more open communications strategy than is common in biotech. We hope that scientists, engineers, and builders of all stripes can benefit from watching how we work.
We are building the infrastructure to efficiently hunt for reprogramming interventions that restore function in aged and diseased T cells. This spans both the world of Atoms and the world of Bits.
Our lab has been firing on all cylinders (pipette pistons?). This month we started two key experiments and improved one of our core manufacturing processes:
We initiated our first epigenetic reprogramming screens, testing our core discovery technology and generating initial data to learn representations of reprogramming effects.
It feels like ages, but only nine weeks passed between our very first experiment and these screens. That pace of execution is a testament to the hard work of our team.
Epigenetic maps of aging and disease
Before we can develop interventions that restore cell function by epigenetic reprogramming, we first need to build a map of the epigenetic changes that arise with aging and disease.
These maps will help us evaluate whether a reprogramming intervention restores a healthy epigenetic state, which we will use as a top-of-funnel indicator for whether the intervention is likely to restore function.
This month, we started our first mapping experiments to chart the epigenetic course of immunosenescence. When completed, we believe this experiment with increase the data available for interrogating the epigenetic basis of immunosenescence by several fold.
10X manufacturing efficiency
A key step in our discovery process involves building a “library” of reprogramming factor perturbations that we can deliver to aged and diseased cells for testing. Manufacturing these libraries is a challenge!
This month, we 10X’d the manufacturing efficiency of our initial reprogramming libraries, allowing us to run more experiment, more quickly.
NewLimit is an atoms-to-bits and back again company. Alongside progress in our wetlab, our computational laboratory has been making GPU fans spin.
Learning representations of lymphocyte differentiation
T cell function is largely determined by the differentiation state of a cell. Remarkably, different T cell subtypes are as epigenetically distinct as other hematopoietic cell types are from one another!
Therefore, we want to measure our reprogramming interventions' influence on T cell differentiation, and this month we built machine learning models that can effectively infer these identity programs from our high-throughput cell profiling assays.
Building a clock, one read at a time
We’re broadly interested in discovering reprogramming interventions that restore cell functions that are lost with age. Some of the functions we’re most interested in are very challenging to measure at scale, like the in vivo performance of a T cell in the context of infection.
It’s tempting to try and train models that can infer these functions from high-dimensional cell profiles, but the functions are so hard to measure that generating training data is non-trivial.
One hypothesis we’re exploring is that training models to infer cell age from high-dimensional profiles — an age “clock” — may be a useful surrogate for these important but tricky phenotypes. Age is much simpler to measure, so we can generate tremendous amounts of data for this proxy task.
This month, we built the first versions of these models that infer cell age from cell genomics profiles and effectively predict the age of new cells from an unseen dataset.