The Next Billion Dollar
AI driven pipelines attempt to prove their efficacy
Throughout biopharma’s history, there have been a number of inflection points, where radically innovative approaches led to the creation of multibillion dollar enterprises. The industry was built on chemical small molecules and for nearly 100 years, most of the pharmaceutical companies had their basis in the chemicals industry. Around the late 1980s there was a realization from companies like Amgen, Vertex and Genentech that you could make large molecules — large proteins — and design them to be therapeutic. That brought antibodies and recombinant proteins to the forefront and led to a new renaissance in the industry. Over the next 20 plus years and up to today, these biologic medicines remain the largest drug categories. Today, there has been an explosion of companies using new modalities from messenger RNA therapeutics to cell and gene therapies, to CRISPR and there is a level of exuberance around the potential these new pipelines hold to cure previously intractable diseases. Similarly, AI and machine learning platform companies are in the process of building pipelines that are likely to create a new generation of industry leaders. The end point is clear, there is a chance to build fully verticalized full stop pharmaceutical companies, that from conception through sales and marketing are driven by automation, machine learning, the integration of interoperable datasets, and where everything is optimized, not just in silos, but using data, software engineering, and machine learning to make every process as smooth and efficient as possible. Companies like Schrödinger and Relay Therapeutics have set the standard, as both achieved successful IPO’s in 2020 and now have market caps in excess of US$3 billion. Other companies such as Atomwise and Recursion Pharma have raised over US$100 million to grow their businesses. “There is no doubt that we will see this grow into possibly a trillion dollar economy as there is nothing more important and demanded by the people then health” said Insilico Medicine Founder & CEO, Alex Zhavoronkov. Investing in biotech is extremely risky due to the regulatory risks involved, the uncertainty around a drug’s efficacy and safety, and the cost of taking a drug candidate to market. What is needed for more companies to break through and become billion dollar companies, is clinical validation. According to NuMedii Founder & CEO Gini Deshpande: “Fully integrated AI companies are not only competing with other AI companies they are competing with biotech companies. Therefore, you have to look at what other biotech companies are doing and ask why are my drugs better than what they can bring to bear?”
“We are going to be evaluated on our drug candidates, very much like a biopharma company.
The fact that our AI platform and technology lead us to the target is fantastic because it is something that other people would not have come up with.
However, at the end of the day, the success metrics are still very much ones that all biotech companies have to face.”
Gini Deshpande, Founder & CEO, NuMedii
Big Pharma’s Role
For AI drug development companies big pharma represents access to capital and validation. For so many of these companies, pharma ends up either being an equity or a joint venture capital provider. That is important, because it funds these companies and it gives them the resources to go and do their missions. Validation is equally essential. If you are a small team of researchers from Boston, San Francisco, New York, London or Toronto, and you are trying to talk to later stage investors about the AI engine you are building, you need some sort of validation, because more often than not later stage investors are not steeped in the sciences. Therefore, having worked with a Merck, Pfizer or Roche, gives companies credibility.
An argument can be made that big pharma companies will have more than enough capacity to build their AI platforms when it becomes apparent that they will play an important role in drug discovery. However, that overestimates their ability to recognize trends early enough to take action.
If you look at the last three decades of drugs that have made it through clinic and have had prolonged success on the market, it has not been the drugs or therapeutics that were discovered internally by big pharma. Instead they were the therapeutics and drugs that were acquired once they showed significant traction, in either preclinical or clinical phase. A report by Stat news revealed that the discovery and early development work was conducted in house for just 10 of Pfizer’s 44 products (23%), while only two of J&J’s 18 leading products (11%) were discovered in house. From big pharma’s perspective, it is prudent to look for companies that have a much more advanced understanding of AI and ML’s role in drug discovery and those companies will likely be acquisition targets as the industry matures.
“Pharma had a similar idea to ours in the 90s and failed because they could not execute. The reason for this is there were not computers powerful enough back then and there was not software that could handle what they were asking it to do. Big companies are good at what they do - supply chain management and sales. Innovating is more of a challenge. You become driven by shareholder value. R&D is expensive and no one wants to take that risk. This is why M&A exists.”
Jill Becker, Co-Founder & CEO, Kebotix
“As an end goal, partnerships offer a certain amount of upside but not all upside. Most AI and data science companies are involved in discovery, and they should think about owning the fruits of that discovery – whether it means owning the IP around a new chemical or biomarkers. I certainly see novel therapeutics in Genialis’ future. We started solving our problem by building software and we were able to monetize that. We then pushed the software into a development context where we have been able to monetize these partnerships. The next big business thrust for us will be moving toward establishing the independence of owning our own assets.”
Rafael Rosengarten, Co-Founder & CEO, Genialis
One of the most exciting aspects of the development of machine learning in life sciences has been that Tinkertoy projects that previously existed solely in academic settings and research labs are now being translated into commercially viable technologies that are able to raise private financing. Leila Pirhaji founded ReviveMed, a company that uses AI to leverage tens of thousands of metabolomic data points to discover novel biology and therapeutics, immediately after completing her PhD from MIT. She is one of several newly minted PhD’s that have spurned joining the ranks of big pharma in order to start companies that commercialize science with enormous potential to fulfill unmet medical needs. The willingness to fund revolutionary ideas did not happen overnight however. Over the last five years the field of AI in drug development and discovery has expanded rapidly and that is a testament to the success and perseverance of companies that pioneered the industry. Auransa’s Pek Lum mentioned: “When we first went to raise money we were told that you can either be a software company, or you can be a drug developer, but you cannot be both. The software driven biotech company was not popular at that time. Now you can see it has become a very crowded space and because of that it is going to drive the ecosystem to a place where it should be accepted.” Companies in the space have become so well funded today there is a saying that - anyone looking to raise capital need only place a .ai at the end of their url and they will be able to raise their series A. Some CEO’s such as Insilico Medicine’s Alex Zhavoronkov, have expressed worry that high valuations given to companies in this field may cause hardship down the road. He points out: “When I see companies from scratch going after an unproven idea, raise hundreds of millions of dollars at valuations that are many times higher than Insilico’s, it makes me worried. In the traditional pharmaceutical paradigm, around 90% of all projects fail. And I am worried that if it happens to many of the overvalued AI companies it will affect companies that worked very hard, invented new technologies, supported the claims with publications, and focused on rigorous experimental validation.” Others have observed that despite the large capital raises, there is now a core group of skilled investors that understands machine learning technology and the upside it can bring to the pharmaceutical field and is willing to stand by companies that possess scientifically sound technology. “Big pharma has seen computer aided discovery come and go over previous hype waves and therefore are skeptical and want to see real proof this time. The quality of diligence and depth of understanding of the best investors is outstanding today. I do not think the companies that truly fake it will survive. You must have something genuinely interesting, insightful, and novel to be successful. I think we are past the peak of the handwaving part of the hype wave and now we are in the proof part. Covid gave companies an interesting chance to show how versatile and real their technologies were by quickly repurposing them to solve an immediate global problem” said Rafael Rosengarten, Co-Founder and CEO of Genialis.
Fortunately, given the similarities to the traditional biotech business model, there is an incredibly fertile, well established equity market that is ripe for building these companies in the private domain. Traditionally biotech companies have between one and three assets and they raise somewhere between US$5-30 million in order to hopefully reach phase one two. After that, typically the company gets acquired by a big pharma company, or goes the IPO route and at that point, early stage investors have made their money. This has proven to be a sustainable model for biotech and it can be equally useful for AI platform drug developers. It might even be more fruitful, as machine learning platforms have a multitude of assets that over time will continue to get stronger as algorithms and computing power improves.
“A company must identify what it is capable of doing that nobody else can do. Because AI technology to some degree is a bit commoditized, you need an exclusive data set, distinct computational capability, or you need an exclusive way of extracting new information from a slide, image or data. If what you are doing is automating something that has already been done, but just doing it with a computer and doing it with AI, that is not proprietary in any way.”
Leo Grady, CEO, Paige
The new wave of computational driven drug development companies are building powerful platforms that they use to generate assets and intellectual property that can be taken to market. However, each company must navigate the tension between putting more resources into their lead asset or continuing to invest in the platform. This creates a dilemma, because If a company invests the majority of its money into a lead asset and divests from its R&D platform, the probability of that asset failing in clinic is still present. Ultimately, it is beholden to the stochastic risk of biology and chemistry and the complexity of human health. Therefore, if it fails, there is a risk of mortgaging the potential of your entire R&D platform on making sure you picked the right asset. Despite this dilemma, billion dollar pharma companies, will not be created simply by deploying software at scale. That is why if you know that your AI works well and it is validated, it is better to start developing your own therapeutic programs because they bring the most value. For this reason, companies like twoXAR, Exscientia, NuMedii, Insilico Medicine and Cyclica all are positioning themselves to build large, and diverse pipelines. The benefit of the proverbial shift from artisan to industrial drug discovery is that these companies will be able to generate several compounds, in contrast to traditional single-asset biotech companies that take an all or nothing approach, where the whole company lives or dies based off of a readout in trials. Companies in the computational biology space will be able to build portfolios, and even use portfolio theory, to think about how to best mitigate risk, just as an investor would. This is the approach Cyclica has taken, as it has a stated goal to create and own 300 programs by 2025. According to Co-Founder and CEO Naheed Kurji “We have seen the benefit of bringing the two models together. Therefore, we continue to work with multinational pharma companies, but we see a profound impact on decentralizing drug discovery by taking our technology and sparking the creation of hundreds of companies and hundreds of programs across therapeutic area. Our model is about aggregating ownership of assets across therapeutic area through partnerships with academic institutions, early stage biotech companies and joint ventures with multinationals.”
At a time when healthcare is desperately in need of new innovation, AI platform companies are generating a lot of excitement. It is becoming increasingly clear that these businesses that were once seen as radical are now becoming mainstream, with robust business models that have the potential to generate billions of dollars of value. They represent a new hybrid way to build a pharmaceutical company that will deliver novel medicines to patients and given the latest news of IPO’s and drugs entering into human clinical trials, those dreams are now one step closer to reality.