5.4.22-Startup-Ecosystem-DeepCure

KFIR SCHREIBER: Hey, everyone, and thank you very much for the opportunity to speak here. My name is Kfir Schreiber I am one of the co-founders and the CEO of DeepCure, and an alumni of the MIT Media Lab. DeepCure is a company we started about four years ago, with a mission to leverage technology in order to discover and develop highly effective therapeutics that are completely unlikely to be discovered today with the traditional way pharma is operating.

And to achieve this mission, we developed, for the past few years, a platform built out of three main pillars that we believe is going to change the way we discover new drugs. The first part of this platform is what is today the biggest database for medicinal chemistry in the world. So typically a drug discovery program starts with a search or a screen.

Traditionally that will be a few millions of compounds and molecules, potential drugs, being explored in order to find a starting point. At DeepCure, we changed that. And today for every drug discovery program, we start with a space of 10 to the power of 18 options. That's one quintillion, about 9 to 10 orders of magnitude larger than any other drug discovery in the world.

But searching a large space is not enough. The big question is, how do we search that. How do we find the perfect drug? And again, traditionally, drug discovery is a very long, sequential process where we try to iterate and optimize one drug property after the other, in order to finally get to a single molecule that will hit all of the different requirements.

But when we looked at that, we realized that we don't want to do this sequentially. The way to do drug discovery is by co-optimizing all of the drug properties together. Think about the efficacy and the safety and the PK properties of a drug. Can we search a 10 to the 18th space, but at the same time co-optimize all of these from the very beginning?

So this is exactly what we are doing by using a suite of machine learning models that we developed at DeepCure. And what it allows us is to first of all go for the drug discovery process in a much, much more efficient way. But more important than that, we can find candidate therapeutics that are not going to be discovered the way we do it today, because they don't necessarily shine at the very beginning. But they have the best combination of properties.

And the last piece is the newest capability we are building today, that we call the Molecular Foundry. So making a drug is not a purely computational effort. You have to go to the lab, synthesize chemicals, test them in a variety of biological assays. And so far we have been doing this the old way, by using human scientists to develop the assays, to synthesize the compounds, and so on.

But today we are building this facility that is a fully automated robotic lab that will be able to synthesize 5,000 unique molecules every month. That's equivalent to about 1,000 human chemists, and generate about 100,000 data points in the same time period, which gives us, again, a huge advantage in the pace of discovery, but also creates this continuous feedback loop, going back to our machine learning models and improving them.

So so far we have been using this technology, this platform, to do our own drug discovery. We have today a pipeline with five oncology programs in different preclinical stages that we are pursuing independently. But at the same time, we now are opening this capability to other partners, trying to leverage this unique technology in order to drive co-development, co-discovery programs with other partners that have unique expertise, whether it's an interesting indication area or unique biology expertise, or anything else that might make a win-win scenario, where we can go together and develop a better drug together.

So please feel free to reach out later on. We'd love to continue the conversation. Thank you.

[APPLAUSE]