The “MMA fighter” of Longevity talks business models, AI drug discovery and working with the right investors.
Dr Alex Zhavoronkov is the CEO and co-founder of Insilico Medicine, the pioneering artificial intelligence company powering innovation in the biopharmaceutical sector. Using next-generation artificial intelligence and deep-learning techniques, Insilico Medicine identifies novel molecular structures with unique properties for drug discovery programmes in different disease areas, as well as for crucial Longevity research.
In early September, the company published a paper in Nature Biotechnology  detailing how it used a new artificial intelligence system to create and validate a series of novel molecules that may be potentially used to treat disorders, such as fibrosis, and certain cancers, in just 21 days — dramatically accelerating the drug discovery process and potentially saving millions of Dollars. This paper is one in series of generative chemistry and generative biology papers, an area Insilico pioneered. A week later, unrelated to the paper the company closed a $37 million funding round  led by Qiming Venture Partners and joined by the nine expert strategic investors in AI and biotechnology, bringing the company’s total disclosed funding to $51.3m.
We sat down with Insilco CEO, Dr Zhavoronkov, to discuss his definition of Longevity and how (with customers and partners ranging from start-ups to big pharma) Insilico’s business model works in AI drug discovery.
Longevity.Technology: You recently announced that you have accelerated the traditional drug discovery time from 3 years to 21 days — what’s your approach?
Alex Zhavoronkov: To understand what we’ve done, you first need to understand how drug discovery and drug development usually works to avoid making exorbitant claims. We show a limited proof of concept of acceleration in just one of the many steps required to make a drug. And the idea behind the paper was to show minimal experimental validation of AI which utilizes the generative and reinforcement learning approaches. Internally, we have much more sophisticated systems that far exceed what we published, and these tools connect the many steps of drug discovery.
Typically, you start with a disease that you want to solve and then you try to find a molecular target. Getting the target right is the most important factor in drug development and discovery, but it’s also the trickiest and the most problematic area. Validation with a target takes a long time, often, several years down the road from the original hypothesis. We’ve accelerated just one of these processes down to two months — shaving-off between eight months to a year from the traditional discovery time. But for many targets that are novel and do not have ready assays the acceleration will not be so significant. Also, there are usually many targets driving the disease.
For the very basic target validation you don’t really need to design a perfect molecule and our AI is able to accelerate the target validation process as well. We also enable target validation with real chemistry that is very likely to go forward into animals and humans. If you look at the entire drug discovery and drug development process, with our AI we are likely to end up accelerating the entire process by maybe two years, with higher chances of getting the target right.
We don’t run our own clinical trials and we don’t currently have plans to take molecules into the pipeline process ourselves — we typically do this with partners. But part of our AI learns on the current and past clinical trials to improve the probability of our pre-clinical programs successfully crossing the finish line. Our team processed almost every clinical trial on the planet. We train our algorithms on this data to make our drug discovery process more efficient and less risky. We try to learn from both failed and successful clinical trials.
Longevity.Technology: Tell us about your recent fundraise and how you plan to use the new funding?
Alex Zhavoronkov: If you look at our investor base, you’ll see several companies specialising in biotech investing — AI and IT. Some of the investors are the top players in chemical synthesis, and biological research , as well as some strategic investors in the longevity space such as Juvenescence. All of them understand how the business and the industry works; they essentially help us set the pace. Often, they don’t want us to go too fast and break things. I could have raised about six times the valuation I got last round if I went with the non-expert investors that only put in the money, but we decided not to go that route because we only want very smart and competent investors who can help us become a better company.
In the longevity space we’ve also got Longevity Vision Fund supporting us. Sinovation is probably one of the most high-profile investors in AI; they’re the most impactful AI fund in the world and have funded multiple unicorns and super-unicorns at seed level.
Most of the investors are interested in Insilico from the drug discovery angle and not because of our focus on longevity. The big biotechnology funds are typically interested in something more established, so they’re not as much interested in Longevity as they are in drug discovery, pharmaceuticals and AI. In our case, we are very lucky to be selling picks and shovels not only to support the gold rush in Longevity, but also supporting the pharmaceutical drug discovery and development.
“In our case, we are very lucky to be selling picks and shovels not only to support the gold rush in Longevity, but also supporting the pharmaceutical drug discovery and development.”
Longevity.Technology: You seem to be working with all sizes of organisations—can you explain your business model to us?
Alex Zhavoronkov: We have two main business models. One is that we collaborate with others to gain revenue; we get immediate and delayed revenue from companies needing small molecule design, or target identification, or both. Those deals are often structured in milestones and royalties — in other words, we participate in somebody else’s success. We also have an internal pipeline of small molecules targeting a variety of diseases. This pipeline is very valuable to us, and the more money we put into this pipeline the more value we can generate by licensing these molecules later to our pharmaceutical partners.
We’re also heavily involved in data economics. We try to have an internal valuation model for data. That’s why sometimes we offer free services, as we get a lot of data in return and can assess the value of this data for drug discovery. We support non-profits, where we can analyse their data, and, our respective goals and agendas align. It’s also why we like to work with start-ups where we can help them generate, analyse, and value the data.
When you work with big pharma, you need to put a lot more effort into partnering, and it can take up to a year to get a contract. Those contracts are not as lucrative as they might seem to everybody else in the industry and you actually do not get access to their data and if you do, you cannot use it. We have very strict data access rules and try not to get access to data whenever there is any privacy risk. There is vast abundance of data in the public domain to train on and develop quality control procedures. We also have a strict rule for working with non-public and non-consented human data. We do not work with the data that is not anonymized. We also do not work with human genomic data. We think that the value of this data for drug discovery and target identification is overrated.
“We think that the value of this data for drug discovery and target identification is overrated.”
With start-ups, we can help them achieve their goals, such as discovering an inhibitor for a specific target, and in return, we get some early-stage pre-clinical data for training our algorithms. Very often the main value we get from a collaboration is the understanding which system or animal model is relevant to human disease. It’s very symbiotic as we help them manage and analyse their data properly and we actually sometimes tell them to generate more data than they need to achieve their goals, just so they can commercialise or use the data later on.
Longevity.Technology: You’re an opinion leader in Longevity — how do you define Longevity to people less familiar with the sector?
Alex Zhavoronkov: It’s easy! Everything that increases the healthy productive lifespan is a Longevity business and I have published on that previously .
I’ve been working in multiple areas of aging research for quite a while now and I’ve published  several works in aging economics and have done lots of work in chemistry and regenerative medicine. I also run a number of conferences and co-organised 7 conferences on aging in Basel. Our group is embracing the mixed martial arts (MMA) principles when it comes to Longevity research. You have to be good in AI, drug discovery, regenerative medicine, gene therapy, and many other areas to extend healthy productive life.
Most of our current work is in chemistry and linking chemistry and biology for pharmaceutical drug discovery and development. But pretty much all this work is Longevity related, as anything I do in cancer, fibrosis, or immunology, for example, will also increase Longevity.”
Longevity.Technology: What disease areas are you working on at the moment?
Alex Zhavornokov: We try to go after disease areas that are implicated in aging and age-related diseases, but we’re not exclusively focused on that area.
In order for us to be more efficient and effective in aging diseases we need to train on something where you have a lot of data available. That’s why we’re going after cancer, fibrosis, cardiovascular disease, muscle wasting, senescence, and immunology. In every one of those areas, there are targets that are implicated in aging.