The StarDrop Story with Matt Segall, Optibrium

Imants Zudans (0:00)

Welcome to Imants Podcast. I am Imants Zudans. I interview founders building tools for drug discovery.

My guest today is Matt Segall, founder of Optibrium. Matt is a physicist, not a chemist. He builds software that helps medicinal chemists make better decisions about which compounds to pursue. There is no founding moment in this story. No garage, no eureka. Matt spent over a decade trying to bring these tools to drug discovery teams. At one company after another he built the technology. And at one company after another, the company chose to do drug discovery instead of selling software. It never worked. Eventually, Matt and his co-founder decided the only way to do it right was to start their own company. That's when StarDrop was born. I talked to founders in this space. They pitched timelines that fit VC fund cycles. 7 years to exit, 10 at most. Matt and his team built a large drug discovery software company. It took 25 years. The story starts at a biotech company called Camitro, where Matt joined in 2001 to build predictive models that nobody believed in yet.

Matt, welcome.

Take me to Camitro.

Matt Segall (1:26)

Thanks very much, Imants. It's great to be here. Well, it feels like a very long time ago now. But it essentially started when I was doing a postdoc at Cambridge University. Although I was a theoretical physicist, I was actually working on modeling drug metabolism using first principles quantum mechanics, so really at the atomistic level to try to understand how cytochrome P450s metabolized potential drug molecules. That led me to join a company called Camitro. It was a Californian based biotech company. It was one of the first generation of the predictive ad-met companies that started up in the late 90s, early 2000s. This is back in 2001, and I was invited to join them out of academia to help to found the research and development group they were building in the United Kingdom. It was actually with a group of scientists who had recently left GlaxoWelcome.

We started working with them on building predictive ad-met models. I was particularly focused on drug metabolism along with the group obviously in California, but we were looking at how we predict a broader range of ad-met properties. Back in those days, what's now absolutely standard in terms of QSAR models for ad-met prediction was a completely new idea, a lot of skepticism about it. It all seemed so simple back then. We were simply going to predict the right molecules, take them and away we would go and have a massive impact on the success rate of drug discovery.

So yeah, that was where we started. And that is where the sort of genesis of StarDrop came from.

Imants Zudans (3:18)

So you said that at that time that was a kind of novel idea, not conventional. How come?

Matt Segall (3:25)

I think just rewind a little bit until sort of the mid-90s, ADME properties were really not even considered until quite late in lead optimization. There were no in vitro ADME studies that were being done in early discovery. Really the first time one would encounter, you know, physical, chemical and ADME properties was typically quite late stage when we started sort of animal studies and looking at the animal models for efficacy. And so in the 90s, a number of the big pharma companies began to recognize that these ADME and PK properties were leading to these late stage expensive failures and trying to think about how to bring the consideration of those properties into early drug discovery and essentially reduce the risk of those failures.

My first boss out of university, a gentleman by the name of Mike Tarbot, was leading a project in what was then GlaxoWelcome, ultimately of course GlaxoSmithKline, called the Combinatorial Lead Optimization Project or CLOP. And it was one of the first projects to develop experimental assays that are now completely standard. Things like microsomal stability or permeability assays that we do very routinely in early drug discovery. That was the first time these higher throughput experimental assays were even being employed in drug discovery. And he took the next step, very innovative, and said, well, if we can measure sufficient number of compounds, then all of this machinery that had previously been focused on predicting potency, so QSAR models, might be applicable to these early ADME properties. And then a number of companies were founded around that concept, Camitro being one of them. There was Lion Biosciences, Ametis, there were a few others.

And these were really the first time that people were trying to build these models at scale and apply them in drug discovery.

Imants Zudans (5:36)

Yes, how does it actually feel to start working on something so new? What were the conversations with scientists that haven't used it before or haven't thought it might be useful? How does it feel actually to present something that now seems obvious?