"We are building one of the first large-scale pharma-tech companies. We use technology in every aspect of our business, from computational drug discovery to tech-enabled clinical trial monitoring."

Why is Roivant’s business model considered to be pioneering in its approach?

First, we are building a next-generation big pharma company structured as a family of companies. Unlike today’s large pharma companies, Roivant is not a single monolithic command-and-control organization. Instead, we are organized as a decentralized family of biotech business units that we call Vants. Our view is that innovation in biopharma most often occurs through small teams. By virtue of our distinctive corporate structure, we are able to recruit top talent and incentivize management teams based off of the individual projects they are actually working on. The second point of differentiation is that we are building one of the first large-scale pharma-tech companies. We use technology in every aspect of our business, from computational drug discovery to tech-enabled clinical trial monitoring. Our embrace of technology is not an end in itself, but rather a means of accomplishing our ultimate goal, which is to develop as many medicines we can for patients as fast as we can.

How has Roivant’s model of in-licensing drugs with early-stage clinical data now shifted to building and designing molecules from scratch?

It has always been the core of who we are to identify biological targets and pathways of interest. We have always done this through interdisciplinary teams that combine data scientists with MD-PhD investor types and translational and development scientists. Historically, when that group of people came up with an idea, such as targeting the neonatal Fc receptor, we would boil the ocean and find drugs at academic centers, biotech companies, and big pharma that matched our hypothesis. We then would in-license, acquire or partner on those therapies. That is still a big part of who we are as a business, however we started to realize that our engine for finding promising targets would sometimes produce a target that we could not acquire. For the most part, up until recently, what we did with those targets was to put them in the discard bin. The discard bin eventually became full, and at the same time some of the data scientists working on target identification made the case that we could do better at using machine learning to design new medicines.

Some of these targets are well situated for the use of machine learning in drug design so we formed a unit called VantAI focused on that problem. We became increasingly interested in this avenue as an additional source of growth, and we have built some really unique capabilities both in terms of computational chemistry and in terms of wetlab medicinal chemistry to make that a reality. That becomes yet another important new source for medicines, so that when we cannot find something to in-license, we can attempt to discover and develop it in house.

What impact do you see computational drug discovery having on the industry, and how will Roivant’s acquisition of Silicon Tx boost the company’s positioning in the space?

Our view is that this is going to be an incredibly impactful set of technologies that could change all aspects of drug discovery. We have chosen to focus specifically on the design of novel small molecules. We are particularly optimistic about targeted protein degraders as a future modality. These are bifunctional small molecules that have a number of interesting properties. For example, you do not need to bind to the active site of a protein in order to degrade it. This is an area of medicine ripe for computational applications. We realized that the machine learning toolkit that we already built in Vant AI is incredibly powerful when you have a lot of data about a system or related systems in order to make predictions. However, sometimes the problem you are trying to solve has no good data out there. In that situation, you want to be able to go back to first principles.

Designing a new computational molecular dynamic simulation from scratch is hard, but we got really lucky in the relationship we built with Silicon Therapeutics. They have what we think is the most precise computational molecular dynamics toolkit out there. Now we can take a new and difficult problem like degrading a tough to hit protein like p300-CBP, and we can simulate that system’s atom-by-atom design using Silicon Therapeutics’ toolkit.