2024 MIT Health Science Forum: Lightning Talk - CellChorus

DAN MEYER: Thank you. Good afternoon, everyone. My name is Dan Meyer. I'm the CEO of CellChorus. And we apply machine learning and artificial intelligence to study how immune cells function and perform. We do this at single-cell resolution and in high throughput for some of the top researchers and biopharmaceutical companies in the world.

The reason we do this is if you are re-engineering cells, recruiting cells, training the immune system, inhibiting immune cells, and you want to understand how your product works, you want to be able to, number one, understand how those cells move and interact because they need to do so to fight disease. And just like every other area of biology, there's wide variation in how those cells perform. So we want to look at a lot of them, but at single-cell resolution.

The most common way that our customers and partners look at immune cell function today is in a bulk live cell imaging assay. We put 10,000 T cells and 10,000 tumor cells, let's say, into a vessel. And then we look at them over time to see how many of the target cells die-- or maybe how many of the immune cells die as well. We don't understand which cells perform best and why, and there's a lot of readouts we don't get.

The good thing is, these are really well-understood assays. The protocols are available, and that's a big benefit to us because we take a similar approach for the user, but we're placing the cells into compartments that have thousands of nano wells at the bottom of them. So an early version of our device is at the bottom-left, and what you see is an individual well with one CAR T cell and three tumor cells.

And when I play the video, you see that this cell, if it plays-- I don't know if we can manually play the video? Maybe not. Thank you. Maybe it didn't come through. So you can go to our website, and I'll have a link at the bottom. You'll see that this cell has to move, form a synapse with-- kill the first cell, and then repeat that process to be a serial-killing T cell. What our algorithms do is identify where the cells are, classify them, follow how they move, and then identify events like cell death, cytokine secretion, et cetera.

The reason we're using the algorithms is we're making thousands of these videos. Every sample we test generates as many videos as Netflix has in its catalog available to you as a user. We're not going to watch all the videos. We're going to use a lot of computer vision AI in order to do that for us.

That gives us, number one, a lot of readouts; number two, they're at single-cell; and number 3, we can filter and just look at cells that kill or don't kill. Just look at cells that form a synapse or don't form a synapse. And then if you're interested in a particular cell or set of cells, we can pick them and do single-cell transcriptional sequencing, et cetera, further down the line.

So, number one, a common question I get is, does it work? We've got data in more than 20 peer-reviewed publications across a range of cell types and applications, et cetera. Number two, does the market value of the data? And we've opened an early access lab where we are getting paid in order to run assays for designing new therapies, selecting lead candidates, quantifying responders and non-responders, evaluating potency and viability of manufactured cells, et cetera.

We're sometimes a little bit of a unique startup. When we talk about customers, they pay for the data. We don't do it for free. And so we do know that people value the data. And importantly, throughout all these applications-- so for early research, for preclinical development and selecting lead candidates, for predicting response in the clinic and doing process development and manufacturing analytics, these are fundamentally the same assay. So people can have data to compare to from early in their process all the way through to process development.

We do have clinical diagnostic and prognostic applications which we're not pursuing now because those have FDA regulations, potentially, and take longer and cost more.

One great example of the platform we published with Kite, MD Anderson, and others this summer where we took clinical product from axi-cel, CAR T cell, as well as clinical development product, and showed that migration and serial-killing is associated with complete response compared to partial and non-responders. It's a short presentation. I would go into that in more detail. Let me know if you would have any questions on that or other examples.

So we're actively running the early access lab and we're about to launch product, working with mostly cell therapy, antibody, and vaccine companies. There are also applications in small molecules and other areas. And we're talking to partners for preferred CRO relationships, instrument imaging, instrument partnerships, and then also international distribution. And if you'd like to see a video, which is normally the best part of my presentation, cellchorus.com/videos or follow the QR code.

[APPLAUSE]