In Suzhou, a city 30 minutes by train from Shanghai, there’s a nondescript building that looks like it’s covered in Death Star surface tiles. In the middle of the building’s facade is a logo of a square computer chip with an Erlenmeyer flask in the middle. The logo seems to glow the color of green battery-charging lights.
A virtual walkthrough starts at the lab entrance. Like a scene from Mission Impossible, the doors are primed to open with facial recognition followed by touch activation on the airlock-like, floor-to-ceiling door. If the video weren’t on mute, I wouldn’t be surprised to hear hydraulic and whooshing sounds.
“I want to make you feel like you’re in Star Wars,” Alex Zhavoronkov, co-founder and co-CEO of Insilico Medicine, told me while showing real-life footage of his fully automated laboratory from his laptop.
As the camera guided me through the futuristic doors of his real-life sci-fi lab—Zhavoronkov has dubbed it Life Star—it pans into a glass room the size of a tennis court, with mechanical arms swinging and autonomous mobile robots shuttling around. “Here the magic happens,” smiled Zhavoronkov.
The robot works with various types of samples—cells, tissues, or organoids—prepping them for various kinds of imaging and next-generation sequencing techniques, including those for profiling DNA, RNA, and methylation. After the robotic ballet finishes generating data, the data goes back to Insilico’s trademark artificial intelligence (AI) for target selection. If anything interesting shows up, scientists will begin a validation process.
Life Star is functional around the clock. And every one of Insilico Medicine’s programs is now supported in some way by the autonomous, AI-powered robotics laboratory. Construction for his robotic laboratory in China began during the COVID-19 pandemic. “The government allowed us to bubble there, so people lived there!” Zhavoronkhov said, showing me a picture of himself lying on what looked like a gurney in a room covered with plastic. “I slept there for four months.”
Zhavoronkov, whose real name is Aleksandrs Zavoronkovs (according to his Twitter), is unabashedly ambitious and brilliant; he’s a borderline mad scientist. He doesn’t fit the sci-fi motif of an evil genius that wants to take over the world (although Insilico Medicine has sites in Suzhou, Shanghai, Hong Kong, Taipei, Abu Dhabi, New York, San Francisco, and Montreal).
Under his leadership, Insilico has raised a total of $401.3M in funding over 10 rounds. Their latest funding was raised on Aug 10, 2022, from a second Series D round led by Prosperity7 Ventures and Aramco Ventures, bringing the total Series D financing to $95 million. New investors, including a large, diversified asset management firm on the U.S. West Coast and BHR Partners joined the round, along with current investors, including lead investor of Series C financing round Warburg Pincus, B Capital Group, Qiming Venture Partners, BOLD Capital Partners, and Pavilion Capital. Zhavoronkov’s fundraising has allowed him to take Insilico global, having opened several R&D centers around the world, and partnered with multiple pharmaceutical, biotechnology, and academic institutions.
Zhavoronkov holds two bachelor’s degrees from Queen’s University, a Master’s degree in biotechnology from John Hopkins University, and a PhD in physics and mathematics from Moscow State University. Since 2012, he has published over 150 peer-reviewed research papers, and two books including “The Ageless Generation: How Biomedical Advances Will Transform the Global Economy.” Zhavoronkov is an adjunct professor of artificial intelligence at the Buck Institute for Research on Aging. In addition to serving on the advisory or editorial boards of various journals and co-chairing the annual Aging Research and Drug Discovery conference, Zhavoronkov writes articles for Forbes in his free time.
The real McCoy
In the past year, Insilico nominated a handful of preclinical candidates, and generated positive topline Phase I data in human clinical trials with an AI-discovered novel target and AI-designed novel molecule for idiopathic pulmonary fibrosis that received Orphan Drug Designation from the FDA and is nearing Phase II clinical trials. Insilico also recently announced that its generative AI-designed drug for COVID-19 and related variants have been approved for clinical trials, as has the company’s USP1 inhibitor for the treatment of patients with solid tumors.
This past year, Insilico Medicine has used its end-to-end AI-backed drug discovery and development pipeline to nominate nine preclinical candidates in small molecules, which Zhavoronkov thinks can be pushed up to fifteen a year.
Of these, Insilico Medicine was able to advance through preclinical trials and into Phase I in about one year. To put that in perspective, recent research from McKinsey shows that, over the past decade, the average time to take a new medication from candidate nomination for preclinical testing to first-in-human trials has been about 26 months. Additionally, for a pharma company seeking to move three to five investigational new drugs into first-in-human studies every year, an acceleration down to a year, applied across the portfolio, could translate into a risk-adjusted net present value exceeding $400 million.
Zhavoronkov, who is very matter-of-fact, has a bone to pick with most people who say they run AI companies for drug discovery and development. “The company should be using deep learning technology to some extent,” said Zhavoronkov. “For me to consider somebody as an AI company, I would really need to see a substantial AI component, like deep learning and generative AI, and the company should at least to some extent contribute to the development of the software field, not just use somebody else’s tools.”
And then there are those who, Zhavoronkov said, claim to be AI companies, but they are just users and those who are doing high-level googling or Excel with advanced algorithms. “I have never seen an incubator company produce a genuine AI product from an AI system,” said Zhavoronkov.
“I’m not going to name them, but usually the founder of these companies was actually not exactly in the field right away.” Successful VC companies like Flagship Pioneering and Foresight Capital were created to access financial markets, he said. “Some of [these companies] missed the opportunity to list because 2021 was the year of abundance. Everybody got enormous amounts of money and huge valuations. Some companies jumped onto the market prematurely, some timed it extremely well, and some did not list. We’re in 2023! Show me a single asset! Lots of hype, but no results. So, the industry has consolidated.”
Zhavoronkov thinks that the resulting environment has been tough on young entrepreneurs who are getting into the field and have good ideas and algorithms, but cannot fundraise because investors have seen their money go up in flames.
“Drug discovery programs are usually like $50 million to scratch the surface; I was shocked,” said Zhavoronkov. “So you raise $400 million and burn half of that in the first year on multiple programs. If you create something great, you can actually generate potential revenue. If you don’t, you’re screwed.”
According to Zhavoronkov, there are only a handful of AI-powered drug discovery and development companies in addition to Insilico Medicine. He cited Recursion and Benevolent AI, although the latter recently announced significant layoffs. If you ask ChatGPT (with data up to 2021) to name two AI-powered drug discovery companies, Insilico Medicine routinely shows up at the top. “I think that we show up there because we publish in this field one or two research papers a month on AI,” he said.
Two are better than one
The reason Insilico Medicine is successful, according to Zhavoronkov, boils down to two reasons. First, they developed a complete end-to-end AI platform in the first few years called PHARMA.AI. This drug discovery engine utilizes millions of data samples and multiple data types to discover signatures of diseases and identify the most promising targets for billions of molecules that already exist or can be generated de novo with preferred sets of parameters. This suite was created to accelerate three areas of drug discovery and development: disease target identification (PandaOmics), generation of novel molecules (Chemistry42), and predicting clinical trial outcomes (inClinico).
Despite the popularity of ChatGPT, Zhavoronkov, not surprisingly, has created something better for anyone asking research questions by text: ChatPandaGPT. This software has integrated advanced AI chat functionality based on recent advances in large language models into its PandaOmics platform. ChatPandaGPT enables researchers to have natural language conversations with the platform and efficiently navigate and analyze large datasets, facilitating the discovery of potential therapeutic targets and biomarkers in a more efficient manner. Insilico Medicine is the first biotech company to implement chat functionality using large language models into its AI drug discovery platform.
Zhavoronkov next shows another video—a sped-up example of using Insilico Medicine’s end-to-end platform. After picking an indication, which Zhavoronkov said is purely based on commercial purposes and can be done with the help of AI, he runs through a use-case of target identification and drug discovery. All it requires is clicking and patience, as it takes hours, even days, to run some of the computations.
Second, Zhavoronkov has a co-CEO, Feng Ren, PhD, whom he said is a real drug hunter. Zhavoronkov met Ren in 2020, while he was at Medicilon, a contract research organization (CRO) providing drug discovery services to biopharmaceutical companies globally. Ren served as Medicilon’s senior vice president and head of the drug R&D service business, with more than 600 chemists and biologists. “He knew how to discover drugs, but he didn’t want to provide services to the others,” said Zhavoronkov. “He wanted to really discover, and he left a lot of money on the table.”
Ren is now using 80% of Insilico Medicine’s resources. “He is utilizing AI to very rapidly accelerate drug discovery and development,” said Zhavoronkov. “That’s the reason why we managed to do nine preclinical candidates last year. We have one proven case where we discovered a new target that generated small molecules and went all the way to Phase I, and Phase II is ready to start in the United States.”
Zhavoronkov wants the best people to work at Insilico Medicine. “We’re super inclusive,” he said. “If you are an alien from a different planet and land in my backyard, come out of this flying saucer and give a hand, I will shake it and say welcome. We don’t care who you are as long as you are really good!”
He first started by hiring people through competitions on AI, such as taking the top three fastest people to outperform the latest release of Google’s DeepMind. With the displacement of people due to the war in Ukraine, at their Abu Dhabi location, Insilico has taken in about 65 “AI refugees” from this region.
Dust in the wind
Part of the legacy that Zhavoronkov wants with Insilico Medicine is that anyone can find drug targets, no matter where they’re from. They could be from countries that have never played a role in drug discovery, or they could be high schoolers. I know the latter case is possible because Zhavoronkov showed me a series of papers that were just published by three high schoolers in collaboration with Insilico Medicine using generative AI to help identify new therapeutic targets for glioblastoma multiforme and aging—his muse.
Like other ambitious biotech entrepreneurs and investors, Zhavoronkov is an aging research aficionado. His interest in aging was his jumping-off point for his entire career. While working a well-paying job in information technology in the early 2000s, he started to keep a pulse on aging research and realized that solving aging would require a computer scientist. So, he quit his job to enroll in the biotechnology program at Johns Hopkins University and then pursued a PhD in biophysics at Russia’s Moscow State University.
Next on Zhavoronkov’s list is to translate all of his work in AI and robotics into a clinical setting. “The patient sample would come in and would be processed in a very similar way, and you would get a prediction for the drugs that were FDA-approved that are good for this patient right there in the hospital,” he said. “If I do this in the next couple of years, I will be able to put a checkmark next to my life.”
While our interview was booked for 30 minutes, Zhavoronkov suddenly paused after an hour of talking. I don’t know why he stopped, although I did spy a notification that said to take a few minutes to himself. I probably couldn’t have stopped him earlier if I had tried. Zhavoronkov is a man on a mission who does not switch off. And I don’t think anyone can stop him.