The hardship of Drug Research and Development

…And then US president Bill Clinton said:

“We are here to celebrate the completion of the first survey of the entire human genome. Without a doubt, this is the most important, most wondrous map ever produced by humankind.”

Then money poured into Amgen, Genentech and all the new biotech firms. Big pharmaceuticals and Universities were having a field day. Phrases like “a brave new world” and “a healthcare utopia” were all what people thought about. The belief in scientific progress and the optimism were shown, through all the green pixels on the Stock market exchange digital screen. The harsh reality is, from 1999 to 2001, only 25% of biotech firms doing IPO had any products in clinical development. Fifteen years later, by 2014 there were only 20% of all biotech firms ever existed had any products on the market or sold to pharmaceutical companies.

The truth is, the general public buy stocks from Eli Lily and Human Genome Science (or, then, just any name that has the word “Gene” in it), hoping to catch the next Amgen and Genentech, the same way they hope to catch the next Google or Microsoft. They bet on innovation, not knowing the underlying differences in development processes, at the same time blowing up the ever-growing Genomics bubble. It is crucial to understand, how Drug Research and Development (R&D) is fundamentally different and infinitely harder than other technological advances.

1. Uncertainty

Drug research and development are riddled with uncertainty and risk. Of course, uncertainty is the reality of all research and development. However, in most industries, mere basic technological feasibility is not at stake in the R&D process. When a Tesla engineer hope to develop a new engine for the new Model S, she can design a hundred different variations of the same motor, and decide on the most esthetically pleasant and economically feasible, but she won’t have to worry about if the final product worked. Whatever design she chooses to bring to manufacturers, she would always have a car that can move from point A to point B.

That, however, is not the reality for drugs. Vast majority of R&D projects fail. “Only one in over six thousand compounds synthesized will ever be approved.” (Pisano, 2006) And that is if the compound worked well enough to be submitted to the FDA. All the screenings, researches and clinical trials had already eliminated 92% of molecules that could potentially be developed into drugs.

This is due to the very nature of drug design. In other industries, development is the process of “imaging and testing a series of manipulations” to existing technology, to achieve some desire functional or economical. The programming languages are there if you want to code a software. The wires and chips are there if you want to build a phone. But drug R&D is the process of identifying and validating possible therapeutic process, aka finding to lock to a disease, then develop hormones, enzymes and other proteins that might cure and lessen the effect of the disease, aka finding the key for that lock. Both of these elements, the lock and the key, are not developed or designed; they are given by nature. For target receptors, scientist have to constantly research and learn new things about the human body, microbiota, genomics, cell biology, … all the while testing and examining if they have therapeutic potential. For molecules that binds to the receptors, scientists have to screen through the entire repertory of knowledge from hundreds scientific fields, at the same time, consider developing the safest and most efficient molecule for the job. It’s true that we know about the building blocks to these molecules: we know atoms and electrons, as much as we know about semiconductors and properties of electric currents. But we don’t know enough, or anything, about millions of genes and enzymes and bacteria and proteins. To put it simply, given today’s limits of biological knowledge and constraint imposed by human biology, drug R&D would be like asking someone program a 11-million-lines-of-code software without using algorithm or having not heard of the concept of algorithm; or like asking someone in the 1700s to design a rotor before the days of Michael Faraday. The languages and electric currents existed then, but we just wouldn’t know what to do with it.

There’s also a lack of testing models for how a molecule would bind with a receptor. Since humankind has known all (or most) there is to know about aerodynamics, it would be easy to build a program to predict fairly precisely how a Boeing would fly at a specific height, wind speed and air pressure. But with what little we know about biology, plus the fundamentally differences in biology between humans, fruit flies and mice, (and unfortunately, it’s illegal to conduct clinical trials on human now), it’s relatively hard to build an accurate testing models for drug R&D.

2. Integrality

The best final products are always simple, neatly designed and smoothly ran. An iPhone with all its intricate chips and codes and Tylenol with all its complex molecules and bonds, both solved problems, calling people and headaches, respectively. Because at the end, in the users’ hands (or stomach), all of a solution’s components must work perfectly, with many moving interdependent parts. Here lies the bane of making a drug: while the iPhone is modular, the drug is integral.

Indeed, the computer is a perfect example of a modular machine. When a computer breaks down, the problem could be its programming, or its graphic card, or memory, or hard disk. A problem can be broken down into the malfunctioning of its parts is called to have modularity. And since the problems are independent of each other, the solution can also be a set of independent tasks addressing different sub-problems, performed in multiple locations and and at different points in time. To code a program, you can have an army of coders around the world to contribute to the project, without affecting each others’ task much (open-source projects). To build an airplane, you can build 100 engines first and design the wingspan 100 years later. It doesn’t make much sense economically, but we’ll still get planes after a century.

A drug, however, has integrality. “The vast majority of drug consists of an active chemical or biological ingredient and inactive ones needed for formulation” (Pisano, 2006). One might think of them as “module”, but as a principle, “the design of the formulation is always based on the properties of the active molecules. And ultimately, the safety and efficacy of a drug is based on both components. This means a scientist can’t choose to develop a protein that does not work 30 minutes after ingestion, when the drug is absolved only after 31 minutes. Integrality means that when designing a product, the designers must keep in mind all of the other parts of the product, and cooperate with others working on the same project to build a functioning product. It does not help that scientists in this field are isolated islands of expertise, which will no doubt have trouble integrating into their team.

To make the issue more complex, it’s not only that the solution has integrality, but the problems also have integrality. The human body is complex, and nature is complex. It’s hard to simply diagnose, even with modern scientific knowledge, if a person’s depression is the result of his neurological pathways, genetic abnormities or microbiome. And if it was the microbiome, which bacteria do we need to target? And since they exist as a colony of microbes, which microbes should we avoid affecting? When the drug is in the body, how should it affect the brains? Or the liver? If the drug didn’t work, where will the problem be? Is is that the protein didn’t bind onto the receptor? Did it take too long to be absorbed? Is the dose too small? There’s simply no way to look at one aspect of a multifaceted problem, and hope to reach the perfect solution.

Unsurprisingly, drug R&D is infinitely more complicated than R&D in other industries. We simply don’t know enough or have the tools to do things fast and accurately enough to make drugs very well. The drug is too complex, human is too complex, and nature is too complex. Maybe that’s why it takes at least 6 years to get a PhD in biology and 12 years to create a drug, while it only takes 15 weeks of coding boot camp to learn programming and one all-nighter to create Facebook.

Reference:

 “Science Business: The Promise, The Reality, And The Future of Biotech”

By Gary P. Pisano, Harvard Business School Press, 2006.