Is AI Drug Discovery Another Bubble?

Is AI Drug Discovery Another Bubble?

Some Advice for New Startups Entering This Market

A Short Article by Ed Addison

NOTE:??The views expressed here are the reasoned opinions of the author, and not to be confused with market facts.??These views are the sole views of the author and are not endorsed by the author’s business and employment affiliations.

According to McKinsey & Company, the “the AI drug discovery industry has grown significantly over the past decade, fueled by new entrants in the market”. In fact, it claims investment in AI grew from less than $1B in 2015 to more than $7B in 2021.??Such investment is primarily in drug development programs. As the JP Morgan Healthcare week in the second week of January in San Francisco returns to normal after the Covid-19 pandemic, it’s not hard to notice the mushrooming of valuations and new entrants in the AI drug discovery space.??With soaring valuations, a retreating economy, and hundreds of new AI drug discovery companies entering the market, does this point to another bubble???I think not, but the dynamics are not simple.??It seems likely there will be a pullback in valuations, and probable that over 90% of the new emerging ventures will fail.??But for those that survive, there may be a new game in town.??

ASSET VALUES ARE DRIVING VALUATION, NOT AI TECHNICQUES

There are at least three AI drug discovery companies that are now public companies.??These include Benevolent AI, Recursion, and Ex Scientia.??Spot checking their recent closing market prices, their enterprise values are all above $500M (and were once all above $1B).??Each of them raised more than $250 M prior to going public.?The valuation driver is asset values, not AI algorithms.??In each case, there is a pipeline of multiple assets with the leading assets at or entering the clinic.??On the other hand, companies with only preclinical assets have valuations an order of magnitude (or multiple orders of magnitude) lower.

Isn’t that the opposite of the promise of AI???AI was supposed to reduce the capital need of drug discovery.??Well, not really.??AI probably does reduce the cost of “drug discovery” substantially.??But it has not yet reduced the cost of clinical development, which is 80% of the cost of drug development.??And since valuations are being driven by asset values rather than AI technology, the capital is still needed for clinical trials.??And the capital is chasing after these deals.

Big pharma is using AI for drug discovery.??These include, but are not limited to Pfizer, Janssen, Sanofi Glaxo, Novartis and UCB.??Most of these initially did so by partnering with startup companies.??However, all of them are building their own AI groups.??Big pharma continues to partner with startup companies when novel approaches are proposed, but not so much for approaches they have already mastered.??The prevailing business model is a partnership around typically multiple drug targets, involving upfront payments, milestone payments, possibly royalties.??The entry point is having a unique approach that has been proven on at least one, ideally more than one, drug target.??The market is data driven and the suitor must have proof data.??This is not a true market, but instead, part of the pharma R&D market.??Deals only get done when something new and promising comes out. Hence, startups should view this as part of their capitalization rather than as part of their ongoing operation.

There has been some hyping in the market by the young ventures by announcing their deal size as the sum-total of all possible milestone payments over the life of all drug projects, assuming 100% success.??This has led to an overstatement of market size by some research firms.??The industry is wise to this now, as the probability of success on these projects is still low.

In observing market conditions, talking with those in the industry and watching he market reports, there is a widespread expectation of a significant pullback in valuations.??Does this mean that AI drug discovery is another bubble?

Further observation shows the entrance into the market of dozens (to more than a hundred) new ventures in the AI drug discovery space.??Based on the usual market dynamics, over 90% will fail.??But something is different now than what the market was prior to the Covid-19 pandemic.

NEW ENTRANTS ARE TACKLING NEW PROBLEMS

In the beginning, almost all the AI drug discovery companies were using AI to design small molecules.??Being investor driven, these companies sought to own all or at least some of the IP.??Small molecule design was a good place to start because it is a well-defined problem.??However, it is a crowded space, because every pharmaceutical company has multiple methods for designing and discovering small molecules.??Further, the molecules have no value, some say negative value, until they reach IND.??Then their value grows enormously.??This explains why the heavy capitalization for those companies that reached the clinic and went public.??It also explains why some companies begin by in-licensing assets at this stage and do not engage in discovery prior.

It is hard to enter the AI drug discovery market for designing small molecules today.??The market is well covered.??The market demands near perfection.??It is a data driven market, and a walk in with a fancy algorithm just doesn’t cut it. You must present what drugs you successfully designed, you must show your data for specific projects, and it must be good, because there are many staid experts in this field, especially in big pharma.??In other words, you must have the goods.??No vaporware or fanfare will suffice.

But the new entrants in today’s market are different.??They are tackling new problems and not the same old small molecule drug design.???Drug discovery for small molecules is a well-worn path, and a difficult one to follow for a new company.??The new problems being tackled include target discovery, modulated biology, polypharmacy, precision medicine, the microbiome, large molecule discovery, drug repurposing, translational medicine approaches, and more.??There seems to be no end to creativity for the new arrivals.??And there is always another AI algorithm that can be tried.

For example, CytoReason offers an AI approach to target discovery.??BullFrogAI uses natural language processing to find late-stage compounds.??And DeepCure is an AI based small molecule discovery company that also applies machine learning to ADMET.??To be sure, there are dozens more.??The new entrants are bringing new methods.??There are new ventures tackling large molecules, biological networks, precision medicine, polypharmacy, drug repurposing, toxicity, clinical trial enrollment, and more.

It is difficult to patent and protect AI.??Instead, the IP sought after is almost always composition of matter and/or use patents for drugs.??The real driver of valuation and deal size is the potential of this IP when the leading asset(s) approach or enter the clinic.??A common strategy used by some startups is to acquire their first clinical stage asset through AI drug repurposing, and then deploy AI for de novo design, attempting to secure a high valuation quickly, and then prove platform worth.??This strategy can work, but it’s tricky, as it introduced the risk of a downstream down round or valuation deflation.??For example, Benevolent AI once had a valuation over $1B, but now it’s just over $500M,, nothing to sneeze at, but declines in valuation can be painful.

WHAT WILL THE LONG-TERM LOOK LIKE – NEW DYNAMICS?

Due to many unmet medical needs, and seemingly no end to creative ideas on how to apply AI to help solve them, this market will continue to expand for a long time.??New startups will continue to enter the market and 10% of the will succeed.??But success will be defined by developing a drug and not by a nifty new AI algorithm.??AI will continue to be applied to the clinical stage of drug development as well, but this space is likely to be dominated by the large CROs.

So, what does this mean for the market???Is it a bubble???As stated, to be sure, there is a valuation pullback occurring.??This is healthy for the market.??But there are new approaches, a seemingly endless supply of new approaches.??After the valuation pullback, we are likely to see a long-term growth period, perhaps more modestly paced, but sustainable.??There is more likely to be a 20-year bull run than a bubble bursting – but after a corrective pullback.??Data Bridge forecasts the AI drug discovery market to be $24.6B by 2029 with a CAGR of 53.3%.

The IPO or exit dynamics are likely to be different, however.??It is harder to become an established leader because those positions have been taken.??Instead, there will be more failures and more acquisitions of new firms by previously established firms.??But acquisitions are great exits for early-stage ventures.??And, therefore, the market will be perpetuated, and it will grow for a long time.

Startups will continue to enter the market with new ideas.??The failure rate will be high, but those who produce clinical stage drug candidates will be successful.??Valuations will be more modest than they have been, and an M&A market will emerge.??But the overall AI drug discovery market will grow for a long time, despite an impending pullback in valuation.??To succeed, a new entrant needs to first prove their method, then book several pharmaceutical deals, and ultimately take one or more drugs of their own to market or to licensing at Phase 3.??????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????

ABOUT THE AUTHOR

Ed Addison is an established leader in the AI Drug Discovery space.??He was the cofounder of one of the earliest AI drug discovery companies – Cloud Pharmaceuticals – where he served as CEO for 10 years and today remains the Chairman of the Board.??With new funding expected in 2023, Ed has had his finger on the pulse of the venture market in this space for a long time.??Ed is helping a large CRO on business development and strategy in AI drug discovery. Ed is also advisor to PolarisQB, a data driven drug discovery company that is based on quantum computing.??Ed advises two venture capital funds that invest in this space.??In his earlier career, Ed was an adjunct professor at Johns Hopkins University where he taught both AI and bioinformatics and was the cofounder of the MS degree in bioinformatics.??Today, Ed Addison is a frequent conference speaker on AI drug discovery and development.??Ed has graduate degrees in electrical and biomedical engineering, and an MBA from Johns Hopkins University and Duke University, respectively, and he attended MIT to study Artificial Intelligence on a sabbatical year.??Ed was twice named “Entrepreneur of the Year” by two different organizations

#AI #drugdiscovery #healthcare #pharmaceutical #biotech #JPMorgan #VC