If you could pick three innovations that would revolutionize drug discovery in the next 10-20 years, what would they be?
I found myself thinking about this question during a recent family vacation to Italy. I was visiting the Galileo Museum, marveling at the state of knowledge during the 1400-1600’s. The debate over planetary orbits seem so obvious now, but the disagreement between church and science led to Galileo’s imprisonment in 1633.
So what is it today that will seem so obvious to our children and grandchildren…and generations beyond? Let me offer a few ideas related to drug discovery, and hope that others will add their own. I am not sure if my ideas are grounded in reality, but that is part of the fun of the game. In addition, “The best way to predict the future is to invent it.”
To start, let me remind readers of this blog that I believe that the three major challenges to efficient drug discovery are picking the right targets, developing the right biomarkers to enable proof-of-concept (POC) studies, and testing therapeutic hypotheses in humans as quickly and safely as possible. Thus, the future needs to address these three challenges.
1. A comprehensive library of therapeutic molecules. My first prediction is that the industry will generate a library of highly selective therapeutic molecules that are safe relative to off-target toxicities. The library will target the majority of proteins in the human body, and will include molecules that disrupt multiple mechanisms of action per protein. Furthermore, the library will be precisely calibrated to quantify the extent of target perturbation, including activation and inhibition. This comprehensive therapeutic library will enable off-the-shelf testing of therapeutic hypotheses very quickly in humans.
Conceptually, this approach is similar to what we think of as “repurposing”. Today, the industry has a relative small library of approved drugs with annotated mechanisms that enable repurposing. These repurposing experiments are powerful when they work, but are relatively limited given the overall scope and complexity of human physiology. Over time, repurposing libraries will expand to include highly specific molecules with known mechanisms of action.
Given the amount of time it takes to develop a safe and effective therapeutic molecule for any one protein and one mechanism of action, I know this seems like a prediction that will take more than 10-20 years. But with advances in computational chemistry, genome editing (e.g., CRISPR-Cas9), and biologics, it is not unreasonable to think that such a therapeutic library will exist in the future. Moreover, it is not unreasonable to think that these libraries will have properties that limit off-target safety and/or the field will be able to test for off-target safety more effectively than we do today (e.g., human organoid culture systems, human microdosing studies that measure gene or protein networks). And when a comprehensive therapeutic library becomes available, it will change how we approach drug discovery.
2. A deep understanding of causal human biology, especially through human genetics. A comprehensive library of therapeutic molecules has limited utility without knowledge of how to deploy that library (or specific molecules within it) in the context of human health and disease (e.g., which disease states should to be treated and how they should be treated). Somehow, causal connections between target perturbation and disease states need to be made. Once causal connections are established, then it is possible to test therapeutic hypotheses in humans. Thus, my second prediction is that the industry will have a deeper understanding of human physiology that establishes causal links between target perturbation and physiological outcomes in humans (aka, “causal human biology”).
While there are many ways to establish causal human biology – see a story in WIRED about immune resistance to Ebola, for example – genetics is one of the fastest growing fields. A July 2015 Nature Genetics article by Matt Nelson and colleagues demonstrates the success of such an approach (see here, International Business Times commentary here). Within the next 10-20 years, it is not unreasonable to think that the majority of the world will have their genomes sequenced. One recent prediction is that the world will have sequenced 2 billion genomes in ~10 years (PLoS Biology article here). If human genetic variation is connected to clinical outcomes of interest – which it almost certainly will through electronic health records, self-reported histories and other sources of phenotype data capture – then it will be possible to create detailed genotype-phenotype maps.
Population genetic theory predicts that most human genes will harbor a series of alleles that perturb function across a range of effects (e.g., from complete loss-of-function to highly accentuated gain-of-function). But it will take millions and millions of genomes – or hundreds of thousands of genomes per disease state (e.g., rheumatoid arthritis, diabetes, atherosclerotic disease) – to establish these connections (see 2014 PNAS paper here).
Genotype-phenotype maps are only the first step towards understanding mechanism of target perturbation. Detailed functional understanding of human genetic variation is required to make sense of human genetic associations. Fortunately, high-throughput biology will likely solve this roadblock. These functional studies are crucial for estimating genotype-phenotype dose-response curves.
There are other molecular tools that will establish causal links between target perturbation via human genetics and physiological outcomes. Mendelian randomization (MR) is one such tool that combines genetic epidemiology with real-life molecular biomarkers. A two-sample study design is feasible today, and genetic and molecular databases will likely continue to expand to enable sophisticated connections. [Note that there is an active social media debate between George Davey Smith and Joe Pickrell on how impactful MR studies will be in the coming years (see Reddit posting here)].
As with comprehensive libraries of therapeutic molecules, it will not be easy to create genotype-phenotype dose-response curves. Still, I predict it will happen. And when it does, the ability to generate therapeutic hypotheses will happen in a matter of weeks not years.
3. Engaged citizen-scientists with real-time monitoring devices and baseline molecular biomarkers to assess therapeutic efficacy. Today, it is a painstaking process to design and execute even a single clinical proof-of-concept (POC) study. There are at least two barriers to fast and efficient clinical POC studies: (a) robust biomarkers that are clinically relevant, and (b) a clinical trial framework that allows for human subjects to be exposed to a therapeutic molecule and then monitored for changes in these biomarkers. Thus, my third prediction is that the industry will have a comprehensive library of human physiological measurements (some might call these biomarkers) that reliably differentiate subtle states of health and disease, and that it will be possible to test therapeutic molecules in humans far more quickly than it is today.
How is this approach different than what is done today? First, the biomarker assays will be directly relevant to physiological outcomes that matter in human health and disease. This is, of course, the aspiration of every biomarker used in clinical trials today. Unfortunately, extant molecular biomarkers are too often divorced from relevant physiological outcomes that matter in human health and disease.
A second, equally important distinction is that these molecular measurements will become part of the daily lives of citizen-scientists. They will be measured through smart personal devices much in the way that we track the number of steps we take each day. The measurements won’t require visits to a physician or a specialized laboratory. The measurements will be woven into the daily fabric of our lives.
Armed with robust and clinically relevant molecular biomarkers, it should be possible to conduct clinical POC studies with off-the-shelf therapeutic molecules guided by causal human biology. In this worldview, real-time patient monitors will transmit molecular information nearly instantaneously. Clinical trial design will be quite different, and a different set of big data statistics will be required to interpret these studies.
It is worth noting that these comprehensive molecular profiles are starting to happen today, through projects such as Google’s Baseline Study and Lee Hood’s scientific wellness company, Arivale. The White House is supporting similar efforts through its Precision Medicine Initiative (see July 8 press release here). Predictably, the data storage requirements for these new genomic datasets will be huge, or as a recent PLoS Biology article described “genomical”, a word play on “astronomical” (see here, here and here).
In summary, the future requires that we identify therapeutic targets with an increased probability of success, select off-the-shelf therapeutic molecules that are proven to be safe in humans, and test therapeutic hypotheses in humans using robust platforms that accurately predict clinical proof-of-concept. If we can rapidly cycle through this process in months rather than decades, then the pharmaceutical field will change forever.
Needless to say, there are many limitations to the approach described above. First, the approach is very reductionist, implicating specific therapeutic modulation as the key to maintaining and restoring health. It may turn out that perturbing complex networks is more important, and that a therapeutic library optimized for highly selective target modulation is not the right library. The concept, however, still holds: rapidly cycling from causal human biology to clinical POC studies. Second, the approach assumes that human proteins are the key to health and disease. If noncoding RNA, epigenetic marks or the microbiome turn out to be more important than proteins, this approach will likely fail. Again, however, the concept still holds, even if the targets themselves are based on a non-reductionist view or non-protein view of causal human biology. Third, there is an assumption that the current therapeutic armamentarium (e.g., small molecules, biologics) will be used to construct a therapeutic library. While this is the model today, I can imagine a world in which we have a very different set of options for therapeutic molecules. And fourth, it may be difficult to engage citizen-scientists for certain life-threatening conditions, especially those for which there is adequate standard-of-care treatment. However, one aspect of this vision is that the clinical POC studies are on molecular readouts that are accurate representations of clinical conditions, thereby preempting a traditional clinical trial. Further, there may be unequal distribution of mobile technologies across the population (e.g., young well-to-do techies may be more accessible than older populations).
Like a Dan Brown novel, some of these ideas require that you suspend disbelief…but why not? (I read Angels and Demons while on vacation, so I am now comfortable with completely unrealistic views on how things happen.) The world will most definitely change, just like it did after the invention of the telescope in the time of Galileo. And that is part of the fun! So please let me know…what three innovations do you think would fundamentally change pharmaceutical development and why?