Liquidity in Drug Discovery

One of the most fundamental problems in drug discovery is that potential drugs don’t have market value until they’ve passed human trials. Unfortunately for prospective drug hunters, running human trials of proposed medicine costs tremendous amounts of money. As a result, only a few players are capable of creating salable assets in drug discovery. The market is essentially discrete; large amounts of money are divided up among perhaps a dozen new approved drugs per year. This situation chokes innovation, since available capital has to be funneled into relatively conservative bets.

The drug discovery market doesn’t fundamentally have to function this way. There are a continuous stream of research papers, appearing in prestigious journals such as Nature and Science, proclaiming research breakthroughs every week. These discoveries might include new mechanisms that underlie critical diseases, or new treatment modalities such as cell-based or nanoparticle therapeutics, or even first steps towards wholesale genetic modifications. Assuming claims are reproducible (not always a guarantee), these papers contain valuable research that launches star academic careers. We could imagine that the stream of basic research launches a healthy flow of downstream drug discovery products which fuel a fluid and dynamic ecosystem. Unfortunately, this is not the case. Even high profile papers typically have very limited market value, and only a limited subset of even the best papers result in the formation of new companies or products.

Why is this so? The fundamental reason is that early research success is often followed by a long development phase which burns capital and doesn’t generate salable assets. Moving forward from an initial research insight, the next significant milestone is a safety study in healthy adult human volunteers (Phase I) study. Crossing from an initial study (with perhaps some animal model validation) to human trials can take years of research, and potentially tens of millions of dollars. Needless to say, this gap can be formidable to cross. In a personal example, despite having spent years doing research in drug discovery, I’ve never had a compound even start this process (the barriers are so formidable that exploratory forays are not feasible). I don’t even know of any users who have succeeded in taking a discovered compound this far. In fact it’s common for many researchers in drug discovery to work their entire careers without ever achieving this milestone. These formidable barriers to forward progress keep the set of companies doing drug discovery small and risk averse. The closest analogues are probably in the semiconductor and battery industries, which suffer from similar capital requirement problems [1]. Just like lithium ion batteries that remain the market standard despite a plethora of alternative battery prototypes, most biotechs and pharmaceuticals have nearly identical discovery strategies and are generally unfriendly to novel techniques.

A number of innovators have attempted to change the pharmaceutical industry in recent years, powered by novel advances in AI and deep learning. These new players usually argue that their partial results (far before human trial success) are worth significant payouts. For example, a number of startups have recently begun to argue that proof in animal models of disease buttressed by computational evidence should be salable assets. Many AI for drug discovery companies have been founded on this conceit. Unfortunately, as these companies have discovered, Pharma companies don’t like buying animal model compounds without human validation. Doing so is risky since animal models may suffer from any variety of degeneracies, and validation is risky and expensive. A few talented entrepreneurs and business people have succeeded in making such animal model sales and deals, but doing so seems to require a formidable connections. A seasoned biotech entrepreneur I spoke to recently said these types of sales require the ability to navigate the behind-the-scene action at big pharmaceutical conferences like J.P Morgan Healthcare [2], a qualification most enterprising drug discoverers lack.

Other recent startups have tried to make large enterprise software sales to pharma. As they discover to their chagrin, large software deals in the pharmaceutical world are perhaps even more challenging to land. The difficulty isn’t in getting an initial trial going; rather, the challenge is in getting any reasonable amount of money for their trouble. Pharmaceutical companies often bundle drug discovery software in with Microsoft Word and expect to pay roughly the same amount of money for each. I know of multiple cases where startups with teams of deep learning engineers have spent years creating meaningful products that big pharma wasn’t willing to buy for 100 thousand dollars. This mismatch between the expenditure needed to fund innovative advances and available sales market has choked innovative companies. Most AI for drug discovery companies are discovering that they need to vertically integrate, and turn themselves into traditional biotech companies (with a veneer of AI frosting) in order to compete in this market. As a result, it is unlikely that any of the current crop of companies will succeed in having truly disruptive impact that changes the fundamental structure of the pharma/biotech ecosystem.

The central ailment for the drug discovery industry is lack of liquidity. Capital isn’t easily available to pay for meaningful advances such as animal model proven compounds or deep learning systems. Why can’t meaningful partial successes like the above receive fair prices from the market? Part of the reason is that tendencies towards vertical integration in the pharmaceutical market mean that companies can’t effectively cooperate to create a marketplace for early leads or improved software. This fundamental market inefficiency needs to be resolved to achieve revolutionary progress in human medicine. I won’t attempt to propose a solution in this short essay, but will depart on the optimistic note that I believe this state of affairs will begin to thaw and shift dramatically over the next few years. As I hinted above, the fundamental problem is one of distrust; pharmaceutical companies can’t effectively cooperate to create a healthy marketplace due to ingrained competitive instincts. Luckily, technology has begun to spur dramatic shifts in the way markets function, and novel technology such as the blockchain enable new paradigms for allowing effective markets to flourish even under the presence of mutual distrust. A blockchain market could one day enable interesting prospects for the future of drug discovery by allowing competitive actors to grow a healthy and liquid drug discovery ecosystem.

Acknowledgements: Thanks to Roger Chen for insightful comments and feedback.

[1] Amusingly though, in a case of the “grass is always greener on the other side,” the battery industry wishes it could be more like the pharmaceutical world.


Written on September 23, 2017