Drug discovery in the age of coronavirus
Biotech Voices is a collection of exclusive opinion editorials from some of the leading voices in biopharma on the biggest industry questions today. Think you have a voice that should be heard? Reach out to senior editors Kyle Blankenship and Amber Tong.
Developing new drugs is incredibly hard. That’s why, despite superhuman efforts from the industry, we’re still looking at 12-18 months minimum before we can realistically hope for a vaccine for Covid-19, and probably months before there’s a proven viable drug treatment.
But our increasing ability to begin to industrialize the drug discovery and development process through an engineering approach means that we have more hope for speeding up this process than ever before — and not just to defeat coronavirus, but to benefit the development of all new medicines in the future.
The traditional drug development process can be broken down into two big “tracks” which have changed very little for decades: prophylactics (like vaccines) which prevent you from getting sick; and therapeutics (like antivirals) that help you get better once you have symptoms. Neither is easy.
For vaccines, that usually looks like first identifying the correct “dead” part of the virus (antigen) so that our immune system can develop the right antibodies; then testing for safety and efficacy (how long do your antibodies last?); then manufacturing at scale (no mean feat; think about all those flu vaccines growing in eggs each year!).
Developing therapeutics is just as hard, requiring a deep knowledge of the underlying biology, including the right target to go after with just the right small molecules or biologics, with high efficacy and low toxicity again demonstrated in clinical trials … and so on. You see why it can take years to understand all of this — sometimes even decades.
But using an engineering approach to developing new drugs with the tools we have coming online today is already transforming this process, making it faster, more efficient and increasing the odds of success. A big part of this is using technology to automate and standardize how we uncover new knowledge about biology — the industrialization of discovery itself.
Biotech companies are doing this by building robotic wet lab experiment pipelines with automation + bioinformatics + data science for rapid measurement and analysis of information in a fully industrialized process. So the sequencing of a virus (now cheap and quick, due to 20 years of advances in sequencing tech) immediately feeds into bioinformatic tools that identify the key parts of the genome; bioinformatic analysis in turn speeds up new ideas for how to target the virus, whether in a vaccine or therapeutic vaccines; new drug candidates are moved into robotic testing massively, and in parallel; and the entire process to human clinical trials is loaded up with more good candidates, faster.
This underlying approach is why Moderna was able to come up with a potential coronavirus vaccine at a speed that blew most industry estimates out of the water. Industrializing discovery like this could work much the same way that factory workforce vs. human speeds things up, standardizes processes, and helps us scale faster and more broadly. It also greatly improves reproducibility, a huge issue in drug discovery experiments when even the way you hold the pipette can affect the nature of the experiment. Now, re-running an experiment starts to look a lot like re-running code — again, easier, faster, and more accurate.
Another critical element of the industrialization of vaccine development is our new ability to use RNA. Instead of giving you part of the viral protein and saying, hey immune system, learn this, an RNA vaccine gives you RNA code (akin to software) for your body to make those viral proteins itself, and then develop antibodies.
Why bother with this RNA middle man? RNA is really, at heart, information, and actually very easy chemically to produce — so this is effectively scaling production by using your own body as the protein production facility instead of a lab making the protein — synthesizing them, expressing them, growing them in, say, eggs for an entire population, all of which is slow and difficult.
If RNA is like software, CRISPR is a whole new hardware platform. Our new ability to edit genetic code through biological design tools like CRISPR is another major vector of attack. For example, one type of CRISPR—CasRx—only goes after RNA: if you give it a guide RNA sequence that has a part of what the virus has, it will “search and find” virus RNA and then cut, i.e. destroy them (teams like Stanley Qi’s are already at work on this).
Now, again, this becomes another bioinformatics problem: Can you identify what the right sequences are? It is also a fundamental shift between those two traditional drug development tracks of vaccine vs. therapeutic: concepts like in vivo blurs the line between both. This you would apply prophylactically like a vaccine, before you get the disease, giving your body a new tool that it didn’t have before to fight the virus when it does encounter it.
In its grandest prophylactic form, this type of technology could potentially address not just previous pandemics, but even future pandemics we haven’t even seen yet. Because in theory, if you did this right, you could identify a sequence that isn’t just for coronavirus, or this year’s flu, but an entire group or category of viruses to “search and destroy.”
Because the RNA sequence covers such a broad spectrum, to evade detection like that, a virus would have to fundamentally change their biology. So the prophylactic treatment already living in us would cover not just Covid-19, but also SARS, and MERS, and maybe even those relatively harmless coronaviruses that cause the common cold.
If one aspect of these approaches is about industrializing discovery and another is about industrializing design tools, is there a way to combine both, and allow us to engineer this process from start to finish? That’s where AI comes in: one of the broad spectrum new tools we have that can industrialize every single stage of drug design. By incorporating genomic analyses from not just the virus at hand but all known viruses, AI can help to identify ideal and potentially novel targets; to identify drugs that can be quickly repurposed; to help come up with new “hits” and lead molecules for novel drugs; for lead optimization of which candidates have the highest potential efficacy and minimal toxicity; even to improve the efficiency of running clinical trials.
As we are learning far too painfully now, developing a new therapeutic or vaccine is not just about accuracy, it’s about speed—and a true matter of life and death. But the good news is, we are finally seeing drug discovery begin to benefit from Moore’s Law. Technology and software tools and mindsets are bringing new forces, tools, and data that will help us speed up and industrialize the development of drug candidates we have to treat a whole host of our diseases — so that maybe, when the next pandemic happens, we can move much more quickly, with much more efficacy … or even eliminate the pandemics of the future.
Vijay Pande is the founding investor of a16z’s bio fund. He is a former professor of Chemistry and professor of Structural Biology at Stanford University where he concurrently directed the biophysics program.
For a look at all Endpoints News coronavirus stories, check out our special news channel.