Five TechBio Myths That Must Go (According to One TechBio CEO)
Preface
It’s been a decade since I ventured out of academia in search of change. While I really enjoyed cancer research and the many wonderful, brilliant people I had the benefit of working with, I was also becoming increasingly weary of the many well-known bureaucratic hurdles that plague academia. This reality would hit especially hard whenever my mentor’s patients, many of whom our lab had the great privilege of meeting, would succumb to brain cancer faster than it would take sometimes for us to even learn if we’d gotten a grant we applied for months prior. And so, armed with enough of that blind ambition and naive optimism that’s typical of youth (or the youthful), I set out to forge a new career path for myself, only this time it would be in tech — an industry I knew little about but had long-admired from afar for it’s famed “move fast and break things” approach to innovation. Interestingly, and unbeknownst to me at the time, just as I began my move from the world of bio to the world of tech, so too was the world of tech beginning to move into the world of bio. By virtues of fate more so than foresight, I eventually came to find myself standing atop the very spot where these two once-disparate worlds had come to coalesce. As someone who’s gone from cancer biologist to VC-backed tech entrepreneur (that’s another story for another time) to now VC-backed TechBio entrepreneur over the past ten years, I truly believe we’re living a historical moment in biotech — the sort that mark timelines as they split the old days of impossibility from the new days of inevitability. This moment is also as much a movement that today has been aptly called “TechBio” (for a deeper dive into TechBio, I recommend this post by Amee Kapadia of Cantos).
By now, it’s no secret that the TechBio movement has fueled a growing number of AI-first drug discovery companies taking aim at the half-trillion dollar pharmaceutical industry: an industry that is inherently data rich but which suffers from historically low drug development success rates. These TechBio companies are of a markedly different phenotype and foundational approach compared to their biotech predecessors, and as such represent more of a necessary compliment rather than contradiction to the many veterans of industry with decades of hard-earned drug development experience in biotech today. I should also point out that I don’t view TechBio as the scientific revolution some seem to profess. Instead, I’m of the firm opinion that TechBio is the exciting inevitability of biotech’s own evolution. However, and speaking from personal experience, even if TechBio is more evolutionary than revolutionary in nature, I also think it’s more of a leap than a step in evolutionary terms. For example, the things we’ve been able to do, ask, and learn at my company (I’ll be able to share more very soon!) would’ve seemed like science fiction to me when I first left academia ten years ago.
My point to all this is that the reality of what TechBio is today is already truly awesome. Let’s not exaggerate it. We’re already developing a (perhaps deserved) reputation for hype and hyperbole when we of all people should know that life-saving innovations cannot become reality without public trust and public market support. In an effort to help bring greater clarity and accountability to how we communicate TechBio’s potential, I figured I’d start by calling out five TechBio myths that I myself can’t stand to see.
Myth #1: It takes $2.5B dollars to develop a new drug.
Less fact than factoid, this oft-quoted problem statement has by now become a pitchdeck cliché thanks to TechBio entrepreneurs (a cliché that I may or may not have contributed to myself at one point).
Although this statement is usually a reference to at least one well-known study (more on this below), what makes it a myth in my book is that it’s so often misused, and by extension, misunderstood. For starters, it doesn’t actually take $2.5B dollars to develop a new drug in most cases. In fact, for many rare and ultra-rare indications, the true cost of developing a small molecule typically falls within the $100M ballpark. What that “$2.5B” price tag so often referenced refers to is the averaged total in terms of R&D dollars spent for every new drug approval that emerges: it is not a reference to the actual cost of R&D for any one drug. These may sound similar but they mean very different things in the same way that, for example, “butt dial” and “booty call” may sound similar but mean very different things. Both cases require one to be very clear on meanings and very aware of appropriate (or inappropriate) contexts.
This issue goes beyond mere semantics too for it speaks of a fundamental misunderstanding of the very issue at hand. As a curious example, consider the fact that whenever folks bring up the ole, “it costs $2.5B dollars to develop a new drug,” trope, they almost always do so while showing a graph of “Eroom’s Law.” In doing so, they’re actually presenting two entirely different studies with entirely different conclusions as if they were one and the same. To clarify, the first would be that 2016 study from Tufts University that placed global research outputs at an averaged $2.5B dollars in R&D expenditures per drug as discussed above. The second refers to Jack Scannell’s 2012 paper in which he famously noted a stark, inverse relationship between our vast technological advances (ie. Moore’s Law) and the growing inability for pharma to produce new drugs (hence “Eroom’s Law”) over the same timeframe.
It’s important to note that what Jack Scannell famously observed wasn’t that the costs of drug development kept increasing, but rather, that the costs kept increasing because of pharma’s decreasing ability to discover new drugs. What makes this observation so striking is that pharma’s efficiency essentially kept decreasing in spite of vast technological advances (and perhaps because of them to some extent). Today, Jack Scannell seems intent on rightfully reminding audiences that pharma’s decreasing productivity is clearly not the result of industry lacking the right technology or sufficiently large datasets to crunch as TechBio often seems to assume. This can be seen in the many TechBio entrepreneurs and investors who seem to miss the inherent irony in proposing to solve this decreasing efficiency problem by *checks notes* throwing even more technology at it.
Myth #2: “Cheaper, Faster, Better” Drug Discovery
If Myth #1 is TechBio’s favorite problem statement, then this would undoubtedly be its favorite solution statement. Indeed, if there’s one thing that unites all TechBio CEO’s, it’s that we’ve all, at some point or another, invariably promised to help make drug discovery somehow, “cheaper, faster, better,” by using our platforms.
To be clear, I don’t think TechBio CEOs are being disingenuous in their claims. Instead, many of these well-intentioned folks (especially those new to drug R&D) are more likely misdiagnosing the efficiency problem as discussed above and, by extension, overestimating TechBio’s role in addressing it. I should know because I too have made such confident claims in the past. Does that mean I now cringe whenever I think about it? Absolutely.
Today, I consider this a quintessential TechBio myth for two main reasons. The first is that TechBio companies don’t, in fact, make drug discovery “Cheaper, Faster, Better” in any significant way as I’ll explain further below. However, for those of you feeling the urge to curse at me through your screens right about now, let me start then by rightly pointing out that some TechBio companies have in fact shown an impressive ability to speed up specific components of the drug R&D cycle. A great example of this in my opinion would be Exscientia’s GTAEXS-617 compound (more recently referred to as EXP617). In a matter of months, they were able to whip up a CDK7 inhibitor with really great looking pharmacodynamics (see below).
Furthermore, and more importantly, this compound appears to be at least as good as Syros’ PhII CDK7 inhibitor, SY-5609, as shown below (props to Exscientia for doing a proper benchmarking analysis — a rarity these days).
This is some impressive chemistry, no doubt, and from a platform perspective a great use case for this would be to enable promising fast-follower programs. This makes Exscientia’s partnership with EQRx a fitting one. Still, we must be fair in that such feats were also largely facilitated by the body of literature that already existed on CDK7. This becomes a lot more limited in diseases where we don’t even know what to target. And, this also largely assumes that the targets we do know of are, in fact, the right ones to even target (rarely the case clinically, unfortunately).
Now that we got that out of the way, let’s revisit why I say most TechBio companies don’t actually make drug discovery faster, or cheaper, or better (at least their current iteration). Firstly, most TechBio advancements are taking place entirely in the pre-clinical setting (or earlier) and any gains there, while welcome, are still trivial in the grand scheme of things (see figure below). Since this point has already been expertly articulated by Andreas Bender and Isidro Cortés-Ciriano, I won’t dwell on it further. However, I highly recommend anyone interested in TechBio to read their 2-part series on the matter.
It’s also important to point out that, as of now, TechBio is actually looking to be slower and more expensive than already established biotech practices. To pick on the big guys a bit, consider my fellow TechBio cos. Recursion and Exscientia. Each has raised nearly $1B dollars from private and public markets, and both are ~10 yrs old. Thus far, Recursion has 4 clinical programs (3 in-licensed and 1 in-house)* and Exscientia has 1 clinical program (in-house asset). By comparison, take Day One Pharmaceuticals with 2 clinical-stage assets across 4 clinical programs, EQRx with 7 clinical programs, or ERASCA with 5 clinical programs. Each of these 3 companies has raised $522.5M, $2.4B, and $620M, respectively, with all of these also being at least half as young as either Recursion or Exscientia. Ardent proponents of TechBio may be tempted right about now to point out that there’s inherent economies of scale built into TechBio that will, eventually, make it all worth it. My response to that is that unless we, as TechBio companies, can point out how going from 1M experiments to 1B experiments can have a direct and predictable effect on the number of new clinical trials we’re running or the number of new drugs we’re getting approved, then all we’re showing is diminishing returns, not economies of scale.
The other main reason I consider this a myth is that you don’t need TechBio to make drug discovery cheaper, faster, or better. A great a example of this can be seen in the rare disease space thanks to companies like Perlara. To give credit where credit is due, never once have I seen Perlara’s CEO, Ethan Perlstein, make bombastic claims (or any claims really) about how they were going to change drug discovery with AI or any other frontier technology of your choice. Instead, they’ve stuck to smart science, swift clinical trial designs, and strong patient community engagement. Already, they seem to be generating promising clinical results and have 7 “Cure Maps” established as of this writing. Newer companies such as VibeBio are showing equal promise to soon do the same.
I want to be clear that I have tremendous respect for companies like Recursion and Exscientia. Both have helped blaze a trail for the many younger TechBio cos that have followed (mine included), and in many ways, by being the first they get to bear the brunt of the mistakes we all collectively make as we figure this out. This, however, gives us even more reason to be measured in what we’re promising, and smarter in how we’re delivering.
For example, I’m extremely proud of the clinical programs we’ve constructed at my company given the greatly increased efficiency and anticipated probability of clinical success which, based on internal benchmarking, is the highest I’ve seen in oncology. However, while I’m very excited to share more on that soon, what I promise you won’t hear from me are outlandish claims about how our AI solves every problem. Don’t get me wrong: our use of AI and what it’s helped us learn is freaking awesome. However, not everything is a problem for AI, and it also takes a whole lot more than AI to make a meaningful difference in the clinic. Therefore, I suggest all of us in TechBio cool it with the Daft Punk vibes for now.
Myth #3: TechBio Moats Are Made of Data
I find that many founders and investors in TechBio are under the misguided perception that data is TechBio’s defining moat. If you’re in the business of finding and developing drugs, data is not your moat.
What if your dataset is the largest of its kind? Wouldn’t that make it a moat? That’s not a moat. What if it’s the only dataset of its kind? Also not a moat. OK, but what if your data generation is inherently scalable and transferable leading to compounding value creation over time while also decreasing indication expansion costs? That’s very cool, but still not a moat.
What every scenario above has in common is that they’re differentiators, not moats. Put simply, what type of data you have or how you’re able to analyze it is just what you’ll use to help you (or hurt you) get a drug to market. However, once your drug has entered the market, there is only one moat that matters and it’s called being “best-in-class.” Any TechBio company entering attractive markets best be ready to battle the onslaught of “me-too” competitors (see KRAS example below) rushing behind you, as well as any potential “first-in-class” drugs that may have already beat you to market.
If you’re a small to mid-sized R&D company (the largest TechBio companies today are still in the small-cap category) with an asset lacking clear clinical superiority, don’t expect to compete in any crowded markets. You may be able to get away with a first-in-class though but only because you’re the default best-in-class. If you think I’m exaggerating about the importance of best-in-class, consider the cholesterol-lowering drug Lipitor. You probably know it’s one of the best-selling drugs of all time, but did you know that Lipitor was actually 3rd to market? And if you did know that, could you also name the first two drugs to have entered the market before Lipitor? Yeah, me neither.
For a more TechBio-specific example, let’s consider Recursion again. This is one of today’s most recognizable TechBio companies, and yet, it’s also one of the most misunderstood in my opinion. You’d be surprised by the number of TechBio investors that misunderstand what Recursion’s true moat really is. Usually, these investors will point to the (very impressive) power and scale of Recursion’s phenotypic screening capabilities as their moat, but as discussed above, that’s actually Recursion’s differentiation. Their real moat is in the many rare disease indications that, on their own, are not big enough markets to attract big pharma. The novelty of Recursion’s approach (from a market perspective) is that by pointing their platform and process to many rare disease markets at once, they may be able to turn many otherwise small markets into one very large opportunity for themselves. In other words, their moat isn’t their technology — it’s that in rare disease markets they’re unlikely to face much competition.¹ Recursion’s platform is the differentiated approach that presumably enables them to go after these markets in a theoretically more efficient manner. But, as you may have already deduced, this also means that Recursion’s moat goes away whenever they go after very big, lucrative markets like oncology.
As a thought experiment, consider the fact that any one big pharma company today could easily buy up all the major TechBio companies — discounted or not — if they so wanted to. They don’t, and likely won’t, because they simply don’t need to. Today, all they need to really do is sit back as all the smaller biotech and TechBio cos race to figure out which targets and markets are worth going after. Once an attractive market has been “validated,” you should expect the big pharma cos to descend upon it like Olympian gods ready to subdue the small players currently inhabiting it. For now, it makes sense for them to throw tens or hundreds of millions of dollars (pocket change for them) into TechBio partnerships as a way to get a better look under the hood.
Myth #4: TechBio Can Take Many Shots on Goal*
This myth comes with an asterisk because I do think it could be conditionally true. To explain, I think TechBio platform companies will be afforded the ability to take multiple shots on goal provided their first 1–3 shots deliver at least one major goal. A very successful example of this would be Moderna. Had their Covid-19 vaccines failed as part of their first indication, the resulting public market sentiment regarding their long-term potential could have dealt the company a fatal or near-fatal blow. But, by deliverying a resounding success early on, they’ve essentially bought themselves the right to a few misses should that happen to be the case in their next indications.
In summary, don’t expect public TechBio companies to fare any better than public biotechs if their first clinical efforts end in failure. In fact, they may fare even worse if their market caps were being artificially propped up by unrealistic expectations.
Myth #5: Indication Agnostic Platforms
I’m inherently skeptical of any TechBio platform company claiming to be able to go after any indication because they’re “agnostic.” And, while we’re already here, let’s establish that agnostic does not mean unbiased. Every single experiment and every single data source comes with implicit bias. Whether one recognizes that or not is a different story.
Tangent aside, I’m of the firm impression that, for any platform to be successful, it must be built with a deep understanding of the disease area you’re going after. I think Jorge Conde and Andy Tran of a16z address this point very well in their recent post on “platform-disease fit.”
As the saying goes, a Jack of all trades is a master of none.
Conclusion
It is said that in his final days, a bed-ridden Oscar Wilde felt tormented by the gaudy wallpaper that covered the four walls of his Parisian hotel room. Indomitable even before death, he couldn’t help but quip, “either it goes or I go,” the joke being of course that the wall paper was clearly not going anywhere.
I share this story because I feel as annoyed by these TechBio myths as Wilde seemed to be with his room’s wallpaper. Unlike Wilde, however, I have no intention of leaving the industry or ceasing to exist anytime soon. Therefore, I urge my colleagues in biotech and TechBio to help dispel these myths by standing firm against them.
¹Credit for this clever observation goes to Ron Boger.
*A previous version of this post stated that all four of Recursion’s clinical programs were in-licensed assets. This is incorrect. One clinical asset was actually discovered/created in-house which the current version now accurately states.
