While the early signals of progress for traditional tech companies are fairly well known — product, customers, revenue — biotech companies operate under a very different timeline, metrics, and milestones. Building a company that bridges computer science and biology for drug discovery therefore not only requires a common language between those two worlds, but also means understanding the inflection points that signal value creation — whether for oneself, for other investors, for pharma partners, or for regulators.
For a first-time founder — or a tech founder bringing the tools of engineering to the science of healthcare — learning this language and the associated milestones is critical. Especially since healthcare startups can experience initial success that then slows to a halt once they realize the difficulty of translating their value through well-established processes and criteria. The expectations of the biotech playbook aren’t hard-and-fast rules, and successful companies have “broken” many of them. But just as Picasso first mastered the traditional rules of art before revolutionizing them, aspiring healthcare founders should know the traditional expectations of the biopharma industry before understanding when taking a different path might lead to inspiration. So, below are 16 common pitfalls to avoid when creating therapeutics companies at the intersection of technology and biology.
At first glance, starting as a service seems like the natural way to go for many startups applying computer science to drug discovery. The company has a great new technology that improves drug discovery; and pharma has a huge discovery budget — surely you can create a big business here?! So the company goes and gets a few great logos and pilots with the likes of Genentech, Novartis, and Pfizer. The list keeps growing, and the pilots are small, but surely, eventually, the company will one day get big upfronts and royalties. The problem is… those big deals almost never come. The pilots either end up remaining small service agreements — or are so back-loaded that the company will maybe see a small payout in 15 years if the drug becomes a big commercial success.
The economics of these deals do not sustain a venture-backed business because value creation in therapeutics development comes in the later stages. The current probability of failure for a preclinical compound is 95% — which means the average preclinical asset has very little value. Given these failure rates, proving that a new technology will moves the needle often requires proof points that will take many years to show. Once they realize this, tech companies that still deeply believe in the potential of their company will decide to develop drugs in house (at least to the point that proves out the value). But that has its own set of challenges, and brings us to #2.
When a company makes the decision to develop drugs, they need to make sure they have the team to match. This may seem obvious, but it’s not — it means expanding the team beyond tech founders to include experience bringing a drug from the bench (or computer model) to the clinic. When 23andMe decided to start its own therapeutics effort, for example, the company brought on a former Genentech science boss as its Chief Science Officer and Head of Therapeutic Development to build a team of “drug hunters” that poured over the company’s trove of genomic data.
Such hires with bench-to-clinic experience have often earned that experience the hard way, through past failures, which means they can more easily and quickly answer questions that traveling through a tech-centric idea maze alone can’t: What is a commercially interesting/ viable area for the company? Is the data good enough? How do we design a great clinical trial here? What questions are relevant for the technology to even answer? It’s important to note that maintaining the technological vision that sparked the company in the first place also remains important. So be sure to bring on expert drug hunters that want to make use of technology, otherwise it will be a recipe for culture clash (and the wrong kind of disruption).
First-time founders in this space often hope they can raise a big round with biotech investors based solely on a technology platform, without actually developing assets. In biotech parlance, assets are chemical, biological, or cellular entities that are candidates for further development and potential registration as a new regulatorily-approved treatment. But despite the real excitement around AI-driven drug discovery and other areas, the vast majority of biotech industry partnerships and venture partner rounds still have at least some assets in their pipeline. If startups do raise significant amounts prior to developing assets, it will typically be because the company is based on very well-understood biology, comes with a leadership team of biotech insiders, or was incubated within an established biotech venture firm.
For a team relatively new to the industry, however, assets are what provide the much-needed validation that the technology “is working”. Perhaps even more importantly, potential biopharma partners as well as biotech VCs are accustomed to seeing such assets. Even if they fully appreciate the underlying technology and resulting data, they already know how to value a company with assets.
For a company to create useful drugs, its in silico predictions — predictions made by computers — must eventually be proven out in biology, too. Furthermore, to understand the risk of an asset, potential pharma partners will want to know the primary biological processes that the drug affects. This is so they can better understand and help mitigate the risk around safety and efficacy that otherwise leads to extremely expensive drug failures in the clinic. Understanding the biology deeply can also help optimize a drug for treatment, find biomarkers that allow better patient selection, and suggest plausible new drug combinations.
Wet lab work (i.e., biochemical or cellular experiments) must therefore validate the technology’s biological predictions and provide the insight that helps answer “how it works” — especially for black box AI techniques — at least until there is robust evidence in humans that these predictions are consistently accurate. For founders, this means putting in the work to understand: (1) the mechanism of action — that is, the specific biochemical interaction by which a drug has its effect; and (2) the drug’s target — the biological molecule that the drug binds in vivo to have its pharmacological effect. Companies not using such target-based drug discovery techniques — for example phenotypic screens (where you try to get a desired effect in an assay or cell with a compound no matter how the compound might work) — most often find themselves in a situation where they don’t understand the biology. They may need to just put in the wet lab work to figure out what targets their molecules bind to, and what they do, by starting with assays or model systems where the biology more clearly translates to a treatment (classic examples are anti-infectives and monogenic diseases, as compared to neuro or oncology indications where cell models are unlikely to translate to such complex and heterogeneous diseases).
Note, in some cases, if the data is good enough you may never end up having to characterize the molecular mechanism — and to dispel a common myth, a defined molecular target is not needed for FDA approval (an estimated 7-18% of approved drugs don’t have one). But this may make some pharma partners uncomfortable.
An old adage in tech is that customers don’t buy technology — they buy solutions to their problems. For pharma partners, investors, and ultimately, insurance companies and patients, a drug is a product that solves a very specific problem. This is the indication: the label that pinpoints which disease and population a drug seeks regulatory approval to treat.
Because first-time founders in this space love to envision their technology as a broadly applicable platform — a very common mindset in tech — they are often late to choosing an indication. Yet having a specific indication is a key part of the go-to-market strategy of a drug asset; it helps frame the package of evidence needed to advance an asset forward. Furthermore, the plan developed around an asset as a whole not only demonstrates the commercial attractiveness of that asset, but also reveals the business savvy of the team: Has the team been thoughtful about market size, unmet need, competition, coverage, tiering, current treatment? Finally, a common mistake here is picking an obviously “attractive” indication and letting the indication guide the biology, rather than having the biology guide the indication. Founders should figure out what the advantage of their platform is, and then pick the indications that play to those strengths instead.
A company could identify an indication with great unmet need, and even deeply understand the biology involved — but that doesn’t mean much if the company can’t also formulate a compound that can become an actual drug. Does your compound have a molecular structure that is compatible with being an effective drug? What might that structure mean for the method of delivery (e.g., pill or injection)? Founders need to know that it gets to where it needs to go; that it will selectively affect the target (not everything else); that it stays long enough to have an effect; and that it leaves when it needs to. To know all this, however, the company needs to demonstrate — with preclinical and pharmacology studies — that it’s tested those parameters, understands them, and knows that they are reliable.
We could write a textbook on what structures make good drugs, and even then a lifetime of experience is what’s most helpful in the real world (though computers can help). Demonstrating the above will at least yield some of the characteristics that make a good treatment, and ideally, the company would be able to show some reasonable chemistry here: a molecular scaffold that has a chance to go the distance as an asset with strong binding; good selectivity; and the absorption, distribution, metabolism, and excretion characteristics that are compatible with a useful drug. Only then does the company know it has the chemical starting point to test its biological hypotheses.
Many companies show up to pitches with in vitro data — obtained from studies in microorganisms, cells, or biological molecules outside the usual biological context — whether in some cells or a biological assay. Generally speaking, such in vitro data shows, in the cell, evidence that their drug does what they think it does.
But the problem is, in vitro data does not necessarily translate to a useful effect in humans. So if that’s all a company has, then it should be incredibly compelling — and have a clear line of translatability from assay to relevant biology that the founders can clearly articulate. A good use of in vitro experiments is to figure out: Can you get delivery into the right part of the cell even in an assay? Does the hypothesis even show signs of working in the relevant context? Can you rule out some of the obvious alternative explanations? The answers to these questions will give partners or investors some initial proof that the asset might plausibly work (which is especially important given the asset failure rates described earlier).
Compared to in vitro data, in vivo data is the place to make a compelling early argument that an asset might have efficacy because it is obtained inside living organisms, usually from animal studies (mostly mice but sometimes larger animals). It provides the compelling data that pushes a company further away from bench to clinic — and closer to pharma partnerships or at least smart funding.
Being able to show in vivo data also demonstrates a company’s expertise around experimental design. How sophisticated is your thinking? Is the study well designed? Do you ask the right questions? Do you design experiments that convincingly answer those questions? How clear and relevant is the effect? This is some of the most compelling evidence you can show to potential partners (short of human data), and great in vivo data can distinguish interesting ideas/ science experiments from bona fide assets worthy of further development.
Toxicity studies test the damage that a new compound can do to cells or living organisms prior to first-in-man clinical testing. Such tox data is critical in evaluating a new asset, and can take many forms — ranging from the in vitro effect of a compound on cells, to the formal 28-day tox studies governed by FDA Good Laboratory Practices guidelines. These are quality guidelines for enabling investigational new drug (IND) applications, which are necessary to bring a drug to human trial.
The successful completion of formal tox studies is a common milestone in tranched funding, where investment or partnership funding might be released upon achieving defined milestones. Sometimes assets or companies will be funded prior to conducting formal 28-day IND-enabling studies, but early animal tox data can still be helpful for understanding more than just the safety of a drug. The goal is to have some hypotheses about what side effects you might see based on what you’ve learned about the biology, and where the drug could have off-target effects based on what you’ve learned about the chemistry and drug metabolism. These studies not only give the company key safety data to advance to human studies, but can help guide how they design clinical trials to mitigate potential side effects.
We’ve cured many mice of cancer and yet are still working to defeat cancer in humans; to state the wildly obvious, mice are not humans, and their biology doesn’t necessarily translate drugs developed on them to human success. So, if a company is aiming to enter the clinic for their asset — rather than partnering or selling it — then a development plan laying out clinical-trial design in humans is key to demonstrating the viability of the product (not to mention the sophistication of management).
Investors and partners don’t want to see that you’ve outsourced this or assumed your partner will do it — they want to see your own thinking and planning. Quality clinical trial design shows off a thoughtful selection of inclusion/exclusion criteria; clinically relevant (and approvable) endpoints; patient recruitment; and a robust design with appropriate statistical power. It shows that the company and founders can recruit the right level of talent to do a very hard thing right. It reveals, in short, a thoughtful company that will have advantages over those that don’t have this thinking developed.
For many tech businesses, value creation is somewhat continuous: You have some revenue. You get some more revenue. That revenue grows over time and so does the business’s value. In biotech, however, value creation is very discontinuous and is tied to even more particular milestones: strong in vivo data, GLP tox, IND, Phase I data, clinical proof of concept. A company won’t get much value from “almost” seeing the data from these milestones (unlike in some pure tech companies where the signals can be more a matter of degree than of kind).
A successful venture-backed company must therefore be able to articulate the sources of value creation, and when they expect them to occur (especially around fundraising events). These milestones could include, among other things, the filing of an IND, a new partnership, or a clinical read-out for a trial. If teams present promises — for example, 9-figure upfront payments based on just a plan or an IND in 12 months even though they don’t yet have a drug hit — they could sink their credibility given those poor assumptions. Companies should show reasonable plans and assumptions about possible deal terms and timelines, because those assumptions reveal a team that knows what it’s doing.
Many tech companies may not need to go as deep in intellectual property because they successfully sell a license to their software (or rely on network effects as the moat for their product). When you interact with pharma, however, you are selling them your intellectual property; your IP, for better or worse, is your product.
Potential partners and investors will therefore focus on both the IP granted to the company, and its freedom to operate. Important IP grant considerations include the strength and breadth of claims, geographic coverage, defensibility against competitors, and others. Freedom to operate, meanwhile, ensures that a company’s IP doesn’t infringe on other existing IP or restrictions. For many companies, this often involves getting a license for a new technology or molecule from a university tech transfer office, at reasonable terms, and making sure it actually protects the asset that’s important to the company. It’s easy to underinvest in IP when you are relying on the development of your data platform, but traditional partners will expect robust IP for both the technology platform and especially the drug assets that come out of it.
Ultimately, startups in this space should avoid under-capitalizing — not having enough money to hit key milestones — and over-capitalizing — raising too much too soon and pushing valuation beyond progress to key value inflection points. But startups still deciding between services-based or therapeutics-based business models often get stuck in between, and therefore don’t raise enough money to show progress on therapeutic milestones. Some of the things these companies can do (relatively) cheaply include: more mouse studies, more in vitro studies, and even good laboratory practice (GLP) tox. Other things, however, are much more capital intensive: running a clinical trial, or marketing a drug.
So what to do then? The capital plan should closely follow the timeline and sources of value creation, and founders should therefore raise enough money to hit the value-creating milestones they’ve promised (with perhaps some cushion). Furthermore, pharma partners and investors will look to make sure that the capital needs behind achieving key milestones (such as IND-enabling studies, IND, Phase I, and clinical proof-of-concept) are reasonable given the company’s plan, indication, and therapeutic modality — they want to know that you understand the key value inflection points, and how to get to them. The other thing to watch out for is raising too much leading to unnecessary dilution if it doesn’t also put new valuable milestones within reach, and it’s generally smart to avoid the funding-comparison trap outside the context of key milestones.
Pharma companies like to deal with people who share the same language, are from the industry, and have shared goals — can you talk the talk and walk the walk? One aspect of this is making good deals that attract quality partners, with economics that are sustainable for the long term business. Signing big-name companies may be great for PR, but if the economics continue to be unsustainable, this will eventually catch up to the company.
A common mistake is expecting that surely your technology will change everything — and therefore expecting everything to change to meet your new tech, as opposed to the other way around (or at least from both ends). The best founders — and corresponding business development teams — can translate the benefits of their technology to the language and practices of their potential partners. So founders will be evaluated on their business savvy here: Can they interact with these companies, attract quality partners, and make big-league deals at commercially viable terms? Traditional biopharma deals often come in several components: upfronts, milestones, and royalties, as well as other more creative arrangements. Startups might start with small amounts of all of these, and the fact of getting them sets important precedent for future deal structures even if the amounts are initially small (the exact amounts can remain confidential). Having shown the ability to get favorable deals, startups can continue to build up the magnitude of these payments and royalties as they build up their credibility, too.
Given the capital intensity of biotech investments, syndicates — groups of venture investors — often join together to fund a company in order to diversify their risk and ensure more capital will be available. These syndicates are not random, though, and can reveal a lot about the company itself: the scientific expertise, further capital sources, connections, other advantages. When potential pharma partners evaluate an opportunity, the quality of investors can be a signal for the strength of the backing pushing the science and the business forward.
The size of a syndicate should also be calibrated to the size of the company; for example, an expensive “big science” platform may require more deep pockets and could benefit by also including crossover investors [investors, often hedge funds or mutual funds, that specialize in funding private rounds prior to an IPO and can continue to support a company by buying shares of its public stock] to validate interest for a company’s future capital needs. In early rounds, single firms may take the large part of the round, and the syndicate could grow over time and subsequent rounds as the company needs to increase the expertise and capital around the table. In any case, founders should always pick and grow their partners based on who can help create the most value for them and their company.
Many traditional biotech investors and strategic pharma partners criticize startups at the intersection of tech and bio for their valuations, sometimes due to initial rounds done by pure tech investors. This becomes an issue when those startups seek additional capital from those investors/partners. The best advice here is that valuation should be a function of progress.
There’s also a common perception from tech entrepreneurs that biotech VCs “only value assets, not platforms”. The reality is that unvalidated platforms have little value… but successful assets provide exactly that validation. Entrepreneurs should therefore be careful that their platform strategy doesn’t inadvertently become a single asset bet if a clinical-stage asset comes to represent the majority of a company’s value due to capital constraints. In that context, a clinical trial failure may mean the failure of the entire company, tossing out a good platform with a bad asset much like tossing the baby out with the bathwater.
* * *
We hope this list can be a resource for tech and first-time startup founders in healthcare who are focused on applying new computational discovery platforms to developing therapeutics in house. Though the biotech industry has traditionally solved the divide between the science and business of developing drugs by bringing in professional management with scientific founders relegated to the scientific advisory board, there is a new breed of founders in the industry. These founders combine expertise at the intersection of bio and tech, and can still take a leading role while understanding and recruiting the expertise of traditional industry players to create new kinds of therapeutics companies.
When it comes to human lives, every drug must pass the unforgiving gauntlet of human clinical trials to become a successful product, and every company will have to navigate many pitfalls to get there. Biology can humble even the best drug hunters out there, let alone the most leading-edge technologists, in their attempts to create therapeutics that save and improve lives. Avoiding many of these common pitfalls may help bridge the gap for startups straddling these worlds of tech and bio therapeutics. And while some technologies may allow traditional steps to be sidestepped, many of the those rules can and should still apply; in this sense, bio is more of an art… and the new masters will emerge from students of the old techniques.
Vijay Pande is the founding general partner of the Bio + Health team at Andreessen Horowitz, focused on the cross-section of biology and computer science.
Jorge Conde is a general partner on the Bio + Health team at Andreessen Horowitz, focused on therapeutics, diagnostics, life sciences tools, and software.