Hidden Gems: Building Pipeline Value with Smart In-Licensing

In-licensing a promising asset can be a quick way to jump-start value, whether starting a company or boosting an existing pipeline. As the theory goes, acquiring a high-value asset that has been derisked can be a particularly cost-effective and low-risk path for drug development.

Making that theory work in practice is a matter of finding the right assets at the right price. Unfortunately, most companies focus their budget on a single in-licensing play, so the effort can quickly become a gamble. Betting on the wrong asset can bake in a failure that burns through millions of dollars (and years of your career!).

?That said, “it’s important to find the right asset” is about as useful as a general piece of advice as “buy low, sell high”, so we’re not going to go there. In this article, we’re going to talk about what makes a great asset, take a critical look at a few different strategies for identifying assets, and then talk about approaches that are more likely to generate value quickly.

The Perfect Asset Is A Dream Home

As someone who’s been on and supported multiple diligence teams in my career, and closed in-licensing deals for several assets, I find the experience to be a lot like house hunting. People start the search with high expectations, and then quickly learn some tough lessons about the realities of the market.

?What makes an asset the “right” one is not where the challenge lies. Everyone wants a molecule that has a strong IP position, compelling clinical proof-of-concept for a large unmet medical need, a derisked toxicity profile, and an easily scaled manufacturing process.

?Unfortunately, the way that diligence efforts are performed, these are often subjective criteria. Much like house hunting, the same asset can be perceived differently, and weighted differently by different individuals on the diligence team. There isn’t an in-licensing package in the known universe that would score highly on all criteria by twelve people on a diligence team. There are always warts to find. And this is not just pickiness. Few molecules – even approved drugs – clean up on every one of these criteria.

?Even if you could find the perfect asset, though, you might not be able to afford it. If a molecule scores high on every in-licensing criterion, it’s going to be reflected in the price. Even molecules that don’t score highly on all of the criteria can end up with inflated prices as a consequence of a bidding war. It’s human nature to value an asset more highly if others are also interested in it. Again, much like in real estate, bidding wars benefit the seller, not the buyer.

?At its heart, in-licensing is arbitrage. The idea is to find assets that have a large upside potential in terms of value creation, do the right experiments and out-license at some point to a downstream partner. Paying top dollar for in-licensed assets – especially when its justified by social proof and not the underlying science – makes that value creation all the more difficult. ?Specifically, the upside potential of the molecule is not impacted by the fact that it’s a ‘trendy’ investment category, while the downside risk remains the same.

Following the latest investment trends when dealmaking is usually a good way to ensure bad outcomes. There’s a good reason why “anticontrarian” investing is not a thing!

Running With The Herd Will Get You Trampled

Of course, when it comes to biotech acquisition, everyone does it. The “strategic” pricing premium that companies are often willing to pay during an acquisition is influenced by the a range of factors, including presence of competitors. Because buyers use the presence of competitors as a proxy for value, a crowded field of buyers can lead to poor choices.

Nowhere is the bidding war problem more prevalent than for clinical assets that have already shown proof of concept. A famous example of this is Pfizer’s 2006 acquisition of Rinat, with a spectacular nine-way bidding war ending in a $500 million price tag for a company whose lead asset had just started Phase II. The price was a record for a privately held biotech at the time, and the asset eventually failed approval due to an unfavorable risk-benefit ratio.

For clinical-stage assets that are licensed prior to commercialization, a significant part of the problem is that late-stage failures in development are fairly common (see the third column in the table below). The risk of late-stage failure is not tested for rigorously at the point of acquisition. (In fact, no risks at all are tested for rigorously in a typical diligence team, more about that in a little bit). Because of this, late-stage asset acquisitions can become speculative in nature.

Okay, so why not just acquire commercial-stage assets then? Sure, if you have the budget for it, but the problem remains pretty much the same. Capturing value is all about being able to open up a gap between what you paid for the asset and what you earned from it. If you don’t have a systematic, data-driven way to estimate the value that you’re going to capture from an acquisition, this (again) is just a trip to the casino.

The past decade has seen some spectacular failures with commercial-stage assets. The lead asset in each of these acquisitions failed to reach its commercial potential.

For example, Amgen’s purchase of Onyx in 2013 is widely regarded as having been a failure. Onyx initially rejected an unsolicited Amgen offer, sparking a bidding war that ended with a $10.4 billion acquisition. Arguably, there was reason to be enthusiastic – the commercial-stage lead asset, Kyprolis (Carfilzomib), had previously shown promising efficacy (compared to standard of care) in a combination trial in multiple myeloma. Kyprolis was expected to reach $2 to $3 billion in 2019, by leveraging Amgen’s extensive clinical trial and marketing capabilities. The commercial modeling looked great, but the technical risk seems to have been ignored during the dealmaking. Things didn’t play out as Amgen expected. ?A critical single-agent efficacy study, months away from completion at the time of the acquisition, flopped, damaging the drug’s commercial potential. An object lesson in the dangers of getting stampeded into overpaying for an asset. Amgen could have saved billions by conducting a careful analysis of the clinical data, rather than focusing on projections of commercial potential that were not linked to drug performance.

Many of the deals in this roll of dubious honor were celebrated when they were made. Even a commercial-stage asset can fail to generate value, if the acquisition fails to meet revenue expectations. Gilead’s 2017 acquisition of Kite at $12 billion created a dilemma, as the small market size for the lead asset, Yescarta (a commercial CAR-T therapy) pushed Gilead towards a high price point. Perhaps unsurprisingly, patients and insurers were sticker-shocked, and Yescarta’s revenues and market share in the years since have fallen far short of expectations.

As with late-stage clinical failures, in many of the cases, these failures were failures of due diligence – all too often, the drug was overpriced or underperformed when it came to safety and efficacy. These are quantitative things, that deserve objective measures – especially when billions of dollars are at stake.

The key take-home here is that acquiring an asset is a business deal, much like buying a house. Whether you decide to buy a shotgun house in the Mississippi Delta for $25K, or a lower Manhattan penthouse with a key to Gramercy Park for $25M, you can get taken to the proverbial cleaners if you don’t know what you’re doing. Not only does a higher price tag not protect you from failure, it increases the risk of a disastrous outcome.

The Shotgun (House) Strategy: The Early-Stage In-Licensing Landscape

Okay, so what about the other end of the market, then? Early-stage in-licensing deals out of academia have been a common pathway for startups looking to jumpstart their pipeline. While some venture firms, such as Flagship Pioneering and PureTech Health, build companies around patents derived from in-house research, most venture firms nucleating asset-based (as opposed to “platform”) startups get the ball rolling with one or more assets in-licensed from academia.

These in-licensing deals are a far cry from clinical-stage acquisitions – from experience, many academic tech transfer offices offer incredibly competitive deals. In 2016, for example, universities took in $2.96 billion from patent licensing, with about $2 billion of that coming from royalties. Yet, in the same year, US drug sales were about $450 billion, meaning that the university revenues were less than 1% of drug sales. Of course, the risk inherent in this approach is far higher. Despite the higher risk, academic in-licensing represents a viable (if protracted) route to clinical success.

But how successful has this approach been, historically? We often hear the claim that most successful drugs originated from the NIH or academic labs. We decided to take a systematic look at the origins of successful drugs ourselves to see what the data says.? According to the FDA’s Orange Book for that year, only about 5% of approved small-molecule drugs can trace the origins back to public sector research institutions (PSRIs). However upon closer examination, the Orange Book misses a lot of well known success stories (e.g. vorinostat (Sloan Kettering and Columbia), valrubicin (Dana Farber) and pemetrexed (Princeton)), as well as drugs originating from non-US colleges, such as the antivirals discovered at the Czech Academy of Sciences). A Nature Reviews Drug Discovery paper from the same era (2010), tells a somewhat different story, assigning 25% of approved drugs (1998-2006, n=252) some level of university involvement in the discovery. About 40 of these were solo university discoveries (15%), while another 20 were co-discovered with either biotech or pharma companies. This suggests that while universities are not by any means the main contributors of innovation, they do in fact play a significant role. This is particularly pronounced in fields like oncology, where academic labs can claim full credit for 32% of small molecules and 40% of biologics.?

While the bulk of new discoveries do still originate in industry (85%), academia does in fact play an important role in drug discovery. This role becomes more pronounced during periods of innovation. For example, during the first 25 years(!) of biologics research in oncology, the vast majority of the assets in pipelines (and eventually approved) were in licensed from academia. The only internal R&D "success" in oncology over this period of time was Mylotarg, which had to be pulled from market for both safety and efficacy reasons in 2010 (more about that later). For many years, the vast majority of oncology biologics originated from universities (or Genentech, which was decades ahead of the competition). ??

That said, even if early-stage assets are cheap, they are massively out of fashion at this point. At some level, this makes sense – 90% of discovery-stage assets fail to make it to the clinic, and of those that do, a further 90% of assets will fail to make it to approval. These are intimidating odds.

?

Why Not Just In-License from Name-Brand Universities?

So, the natural instinct is to look for simple rules to increase one’s odds of success. One obvious go-to would be to zero in on prestigious institutions for sourcing assets. But what makes an institution prestigious? Focusing on simplistic criteria like “Ivy League assets only” wont’ get you very far. Dartmouth and Harvard are more closely matched on the football field than in grant funding. Focusing on name recognition brings to mind institutions like MIT, Rockefeller, and Sloan-Kettering. As we will see, this ‘vibes-based’ approach is about as useful as relying on the merits of an institution’s football team.

A more practically useful way to define prestige is to base it on the level of innovation. Formal rankings of innovative universities have been put together by groups like Nature and Reuters, considering factors like patent productivity, industry collaborations in peer-reviewed journals, and citations. As expected, we see big names like MIT (Reuters: 2, Nature: 3) and Harvard (Reuters: 3, Nature: 33) towards the top of those lists. However, we also see names like the University of Texas (Reuters: 6, Nature: 5) and the University of Wisconsin (Reuters: 25, Nature: 36).

When looking at drugs that were “co-discovered” with industry, some of the same big names popped up repeatedly. Institutions like Harvard and the NIH have had several collaborations lead to FDA approved drugs. This track record of success highlights the value of academic-industry collaborations.??

However, it turns out drug approvals originating solely from academia are not dominated by a few major players but are instead spread out across many innovative institutions. This trend is particularly pronounced in oncology, where only Sloan-Kettering and the Czech Academy of Sciences are reported to have more than one drug originating out of their labs without any ties to industry. Notably missing are “glamorous” names like MGH and the NIH. On the other hand, less flashy names like Indiana University, Tulane University, and Boston University each lay full claim to an FDA approved oncology therapy. This is even more notable in the field of infectious diseases, where you can find game-changing, university-discovered drugs like Emory’s emtricitabine, which is included by the WHO on its List of Essential Medicines.?

So to return to the original question: does academic prestige matter? As it turns out, it all comes down to how you define prestige. While there have been a handful of high-profile approvals originating from the big players (generally in collaboration with industry), the vast majority of academic-driven successes originated from far less flashy institutions.?

Okay. Just Focus On In-Licensing Inventions From World-Famous Scientists, Then?

Of course, you could take the brand-name thing one step further and focus on the very top tier of academic scientists. Many venture firms do exactly this, focusing on a handful of well-known labs that are virtually household names (at least in the Kendall Square area!)

Is this a winning strategy? To look a little more deeply into this, we looked a few years ago at the exits from one such lab (anonymized here). This lab has created over 30 companies over past three decades, including 15 drug development focused companies. That’s quite the record!

Of these 15 drug development companies, five achieved exits or ceased operations, while the other 10 are still active. Out of the five companies that are no longer active, four have had all investment and acquisition deal terms disclosed. Two were moderately successful and the other two were at a loss. These successes generated only mediocre internal rate of returns (IRRs) (20 – 30%), and when pooled together with the failures yields a modest Internal Rate of Return (IRR) of 9.3% over a three decade time-frame. That’s quite a bit short of the 20-30%that VCs and private equity firms seek for their portfolios

Meanwhile the 10 companies that are still active have over the past 15 years raised about $1.5 billion in VC funding and another $750 million through IPOs. With this $2.25 billion in hand, they have produced one FDA approved novel drug and two approved generics, along with another 20 or so drugs still in clinical trials. The approval occurred nine years after founding of the company in question. In comparison, current industry estimates for each approved drug are currently $1.385 billion in out of pocket expenses (plus more in opportunity costs) and a development time of about 12 years. So these active drug companies are performing at roughly the industry averages in terms of cost and efficiency.

Meanwhile the 10 companies that are still active have over the past 15 years raised about $1.5 billion in VC funding and another $750 million through IPOs. With this $2.25 billion in hand, they have produced one FDA approved novel drug and two approved generics, along with another 20 or so drugs still in clinical trials. The approval occurred nine years after founding of the company in question. In comparison, current industry estimates for each approved drug are currently $1.385 billion in out of pocket expenses (plus more in opportunity costs) and a development time of about 12 years. So these active drug companies are performing at roughly the industry averages in terms of cost and efficiency.??

??

Optimism Bias: You Can Go Broke Buying Lottery Tickets

All of this sounds somewhat grim – we discussed that simply looking at the data package is not enough, as the assets that are the most desirable come with a price premium that often exceeds their earning potential. We’ve also seen that there are no strategies for generating market-beating returns that you can implement without digging into the specifics of the asset. Neither the stage of the asset nor its pedigree protect against catastrophic failures. Unfortunately, it gets worse.

Market returns for biotech and pharma assets are extremely fat-tailed – they follow a power law distribution. This means that only a small number of assets generate most of the returns on investment. The commercial upside is front and center in people’s minds at the time of acquisition or in-licensing. Pharma companies pay close attention to the commercial modeling, and VCs base their calculations on the long-run expected rate of companies succeeding (as explained so clearly by Bruce Booth, here).

While this approach makes sense for a VC considering the average rate of returns across a portfolio, using commercial forecasts for decision-making when it comes to a single in-licensing acquisition can be misleading.

These models are often filled with generic assumptions for the entire asset class, rather than for the asset in hand. Even for a commercial-stage asset, there are many asset-specific factors that drive revenue – competitive landscape, legal outcomes for the “patent thicket,” and willingness to pay (by insurers, governments and patients). The forecasts are typically anchored to the total addressable market (patient population X average revenue per patient). But, these numbers can be wildly off-base, most often because market penetration (the percentage of patients that take the drug) is overestimated, or the realized revenue per patient is lower than expected. Focusing on the total addressable market can create glaringly inaccurate projections for commercial-stage assets.

With clinical-stage (or preclinical) assets, there is a second problem that muddies decision-making further. Contrary to popular belief, the upside upon commercialization is not what drives the value realized from pre-approval drugs— it's clinical failure. Around nine out of ten discovery assets fail to reach the clinic, and of those that do, another nine out of ten will fail to reach approval. While calculating the size of the pot of gold at the end of the rainbow can be exciting, it’s the wrong place to focus on investment decisions. There’s a name for this tendency to focus only on the upside while ignoring the low probability of winning – it’s called "optimism bias." This “lottery ticket” mentality stems from an underlying flaw in the typical decision-making process.

Let’s go back, for a minute, to the wish list that teams use to identify a desirable asset. Many of the criteria (popular modality/target, promising efficacy, large unmet medical need, clinically derisked for toxicity) are fairly subjective, and teams will often poll their members for scores during the decision-making process. A score between 1 and 5, averaged across twelve team members is still a subjective, not an objective, measure.

Commercial models, on the other hand, provide a seemingly “hard” number for teams to latch on to (Of course, as mentioned before, most of these models are not even directionally correct.) And it’s exciting to think about the upside!

Weighting the commercial upside too heavily turns the in-licensing evaluation process into a trip to the casino and can feed into the bidding war mindset (which as we saw is a great way to make a loss on an acquisition).

There is a Better Way

As we’ve seen throughout this article, following the herd is not the best way to make a small fortune in biotech (unless you start with a large fortune, as the joke goes).

A simple way to differentiate your in-licensing strategy is to start with a different set of assumptions. (It works even better if your assumption is closer to the truth than the standard assumptions). We discussed above that the assumption that the value proposition of an in-licensed asset derives from the total addressable market is wrong. Since 99% of preclinical assets (and 90% of clinical assets) fail, the expected return from a pre-commercial stage asset can be calculated using a different set of assumptions.

For any given asset, the full distribution of outcomes is not just driven by the upside if it’s commercialized, but also by the probability of failure. This, for example, is what the distribution looks like for a Phase I asset:

Now, as an asset moves through development, its return distribution evolves. Notice that the shape of the power law on the right doesn’t really change, it’s just that the asset becomes increasingly likely to yield returns on the right.

These sketches make a pretty simple point: the further away from commercialization you are when you’re in-licensing an asset, the more likely it is that the asset’s value is being driven by the risk of failure.

Okay, so what? Here’s the thing – the risk of failure is a hard number that can be calculated in an asset-specific way, with preclinical or clinical pharmacology data. In brief, you can use modeling to build a precise quantitative understanding of the therapeutic index (the difference between the efficacious and toxic doses) of the asset that you’re looking to license and leverage that to project performance in later stages of clinical development.

Say you’re looking to in-license a potential best-in-class preclinical-stage oncology drug from a crowded asset class. You can ask the question – is this asset likely to succeed or fail in Phase I? This can be done using translational modeling (the devil is in the details, but) – you can use preclinical xenograft efficacy data and toxicity studies to to predict the clinical therapeutic window. If the asset above was already in Phase I/II, you could ask a similar question, but now you could model the clinical data directly, looking for the answer using powerful model-based techniques such as Bayesian statistics, population PK models, and disease progress modeling. These models can then predict the likelihood that the prospective asset achieves higher efficacy at its projected MTD. By doing this, you’ve turned the fuzzy question of “how do I feel about this asset” into a concrete one. PK/PD projections form the backbone of go/no-go decision-making for asset identification in this approach.

Moneyball, rather than Powerball.

This is a complex topic, and we’re already at least five or ten minutes into an article that you might have been hoping to skim, so I’ll spare you the gory details here. If you’re curious to know more about how this works, we’ve already posted a couple of white papers about how you can use model-based approaches to find “hidden gems”, based on a couple of different investment strategies, namely “asset flipping”, honing in on a drug that has previously failed in clinical development and is available for licensing or re-development, and finding better best-in-class drugs. These are just two examples, but model-based approaches can be generalized for asset identification across a wide range of different investment strategies, including first-in-class assets, very early-stage assets, and commercial assets. The details in each case will vary, but the general idea remains the same.

You might be asking at this point, “but isn’t modeling expensive”? Well, sort of. A robust model-based assessment (using clinical PK/PD or pharmacometrics approaches as we discussed) will run you somewhere in the six figures for a Phase II or III asset. If you decide to in-license that asset without conducting a model-based assessment, you will be out of pocket anywhere between $70 and $400 million on average, for the upfront payment alone. A few months and a few tenths of a percentage point of the money being put at risk in the upfront payment is insurance worth paying for!

The Bottom Line (About The Bottom Line)

If you were to take away a few tips about in-licensing from this article today, they would (hopefully) be:

?1.?? Stay away from land wars in Asia: Bidding wars help the seller, not the buyer. They drive up the cost of asset failure, but don’t make it less likely.

2.?? All in-licensing strategies are vulnerable to mispricing: Focusing on later-stage assets, commercial assets, prestigious universities, or big-name labs all come with a higher relative asking price baked in. There’s no evidence that this higher relative asking price translates to a higher relative likelihood of success.

?3.?? Use quantitative (model-based) approaches to calculate risk: Most of the decision-making process (for example in due diligence teams in Big Pharma) uses metrics that are graded subjectively. Commercial forecasts are also unreliable. Pharmacometric (PK/PD) modeling approaches can be used to calculate the risk of failure in the next stage of development – a key driver of asset value.

?4.?? The hidden gems are the assets that are less likely to fail: In its simplest terms, in-licensing is arbitrage. If you buy an asset that’s already at one value inflection point and successfully advance it past the next value inflection point, you can expect an order-of-magnitude increase in asset valuation. Being able to spot assets with an objectively lower risk of failure (separate from factors that are already priced in, such as later-stage assets or reformulations) can be used to underpin an arbitrage strategy. If you’re asking the question “yes, but how do you create value across a pipeline with this approach”, keep an eye out for the next article in this series, where we’ll dive much deeper into this question.

?Oh yeah, there’s one other thing. (This is LinkedIn, not STAT, after all). We at Fractal Therapeutics have been focused on developing and applying model-based techniques to make drug discovery and development more cost-effective and efficient, every step along the way. If you’re interested to learn more about ways in which model-based approaches can maximize the success of your in-licensing strategy, feel free to drop us a line on LinkedIn or on email!

This article is the first of a series on the topic of asset acquisition and investment (just in time for the J.P. Morgan Healthcare Conference), so stay tuned!

?