How Cellarity's AI/ML Approach Differs

Current drug discovery often focuses on a single molecular target. How does Cellarity's AI/ML approach differ, and what advantages does it offer in terms of uncovering novel treatment opportunities?

At Cellarity, we are harnessing the power of AI and multi-omic data to drive a radically new approach to drug discovery. And what's very specific about our approach, about the Cellarity platform, is that we are shifting the starting point of drug discovery. We are shifting the starting point from a single molecular target to the cellular dysfunction underlying disease, and that shift allows us to really unravel the complexity of this biology and create drug candidates that, in a sense, are out of reach with the traditional method of drug discovery. I mean, for the past decades, drug discovery has barely, barely changed.

It's been a very reductionist process by which you take a complex disease, and you reduce it to a single molecular target, and you place a bet on this target at the very beginning of the drug discovery process. Sometimes it works and produces amazing medications, but most often it doesn't. It doesn't, because, obviously, reducing a complex disease to just one protein or one target, unfortunately, doesn't allow you actually to fully appreciate the complexity of this biology. And very often, things don't translate from test tube to animals, or from animals to humans. At Cellarity, we are focusing on the cell. We have developed unique capabilities that allow us to link biology and chemistry with high dimensional multi-omic data, and that allows us to see biology that other can't, and that's very much what's at the center of what we do.

Cellarity's platform leverages deep learning to identify and reverse cellular dysfunction. Can you elaborate on the specific types of data used to train these models, and how this data informs the creation of non-intuitive treatments?

I mean everything we do starts with healthy and diseased tissue, and we collect the transcriptomic profile of all the cells in these tissues. So being able to collect high quality bio samples is absolutely paramount, and we are actually we have developed foundational model that are based on different types of data, mostly single cell transcriptomic. We can use bulk transcriptomic or single cell nuclei data as well, and that's very much, actually the foundation of our platform. Now we can add all the additional data types to this transcriptomic data. We can add proteomics. We are adding, actually, very often, ataxic, chromatic chromatin accessibility, and that allows us, obviously, this is more fit for purpose data to really understand biology more precisely for a specific program. That's how our platform operates.