Museum of Natural History. That’s a funny name, isn’t it? The whole building is full of scientific and biological items like skeletons and fossils. Where’s the history part? Shouldn’t it be called Museum of Science? Most people would put science and history on the opposite end of the spectrum.
When you hear a name like that, it almost feels like someone’s trying to obfuscate, trying to make something sound like another thing altogether, don’t you think?
Sometimes that’s the case–the name is meant to divert you from the true meaning, like “fructose,” a term meant to make the sugar sound more natural by making it sound like it comes from fruit. But in other cases, the name is accurate, and when you dig into it, you find that the name reveals an underlying truth.
“Natural history” is a term for biology and similar fields like geology/geography (Yes I know people don’t put geography and biology in the same bucket but hold on for a sec). It is distinguished from “natural philosophy” such as physics and chemistry. It’s not philosophy because philosophy is timeless, based on first principles. It’s history because accident of evolution have produced the organisms we have on Earth. The organisms we will find on another planet are not going to be the same as the ones on earth, while physics and chemistry will be the same.
For a long time, chemists and physicists looked down on biology. They called it “stamp collecting.” And indeed, most of what biologists used to do was collect samples of plants and animals and categorize them. Genetics was the first true scientific contribution that biologists came up with, soon followed by the theory of evolution. But true “science,” as in unyielding laws that is universal is still sparse compared to physics and chemistry. Evolution is still likely to be true on other planets, for example, but molecular biology like DNA as the carrier of genetic information is almost certain to be quite different.
So biology is the study of history of organisms as they happened to have occurred on Earth. Many things in biology are impossible to explain without referring to the past. The way we develop as embryos, for example. Why do humans have gills at one stage as embryos, for example? Only because in our evolutionary history, we were fishes once. Why do American antelopes run so fast? Because they had to escape the now-extinct American cheetahs.
We as biologists have moved from the first stage–namely stamp collecting, to the second stage–namely pattern recognition, and have started to move into the third stage–namely universal laws. These are stages all fields go through. But we have a lot more progress we need to make on the universal laws, before biology can become natural philosophy.
But, regardless, biology is a science. Biologists are scientists. We follow the scientific method, which most people consider to be the hallmark of science. We like to contrast science and humanities as the two ends of a spectrum. But science and humanities share something in common, which is why both are considered liberal arts. Both pursue the understanding of the truth. The Platonic ideal. Both try to make sense of the world. Both trade in the currency of ideas and theory.
The true opposite of science is not humanities. It’s engineering.
Engineers are concerned with practical things. They build things–buildings, bridges, software. They’re solution-oriented.
The problem is, in drug development, we’re trying to build things–namely, drugs. And scientists have never been trained to build things, only to try to understand things. There is a big difference.
In biology, we still have very limited understanding of how organisms work. We’ve come a long way, but I would be shocked if it turned out we today understood more than 1% of biology. Or to put it another way, I suspect that in 50 years, we will know 100 or 1,000 times more about biology than we do today. And on top of that, at least half of what we believe about biology today will likely turn out to not be quite right.
As scientists, though, we are not trained about the limits of knowledge. If we see three dots, we draw the best shape we can around the dots, and then act as if that’s reality. We haven’t discovered the hundred other dots, but we draw a triangle and the world is a triangle until we discover the fourth dot, at which point the world becomes a rectangle.
And that’s fine, because it’s better than imagining other dots that may not even exist, which is the alternative. We’ve made tremendous progress in understanding the world. If we’re a pale version of what is to come in the next hundred years, we’re a bright supernova compared to the superstitions of two hundred years ago.
Except when scientists try to build things. We think we understand enzymes and kinases and we develop drugs to block those enzymes. And after two hundred million dollars, we find that it doesn’t work. We think we understand Alzheimer’s disease and develop an antibody to bind to beta amyloid. And after half a billion dollars, we find failure.
We scientists in drug development approach our goals like we’re building Webvan.com instead of approaching it the way modern internet companies do, with rapid prototyping and empiricism (you can read more about that in “Lean Startup”. We have a grand theory and we spend a fortune in our belief. We have a strong undercurrent of disdain for empiricism. We shudder at the thought of developing a drug without a mechanism of action (never mind that the true mechanism of action not infrequently turns out to be different from what was originally thought to be the MOA).
We’re scientists. We are reductionists. We want to understand every step in the pathway. We love many narrative fallacies. That one signal molecule binds to one receptor. That biological processes proceed in a linear stepwise fashion, with one molecule triggering another, like gears in a watch. We love cascades, like the clotting cascade and the complement cascade.
We don’t like system thinking. We don’t worry about feedback loops. We don’t understand harmonics and resonance in biological signaling networks. We recoil at the thought that biological systems may have nonlinear response to interventions, or that genetic background can profoundly influence effect of single gene mutations.
As an example, let’s look at the inflammatory cascades. Complement cascade, or the cox cascade. We learn in school how those cascades are activated. But we don’t learn how they are deactivated. Or more importantly, how the deactivation is controlled and modulated. How does the body know when its time to shut off the inflammation, and how is that done? If you were to ask a biologist, you would get blank stares, or they would say, “well, it just peters out once the stimulus is gone… I think…”
No. Almost all biological process are actively modulated and actively quenched. Almost no biological process “peters out.” But that’s not how biologists think. That’s in contrast to engineers. When I talk to electrical engineers about biological cascades, they immediately zoom in on feedback mechanisms. They tell me that the most important part of any network is the feedback loop. I tend to agree.
This has direct consequences on drug development. Most anti-inflammatory drugs are designed to block some part of the activation of the inflammatory cascade. Very few are designed to enhance the quenching of the cascade. Same with most other indications. Most cancer drugs are designed to block the activation of the growth signaling.
The problem is that often, blocking one part of the cascade doesn’t work. In fact, it often produces the opposite of the intended effect, because the feedback loops in the system only rev up the cascade in response.
This is not to say we don’t see a few instance of the more systems-oriented therapies. Harnessing the ubiquinase system, like Kymera is doing, for example, is a fantastic approach.
And, to add insult, our tools are scientific tools, not engineering tools. We use genetically homogeneous mice for our experiments for example, which are great at answering subtle questions about biology, but terrible for drug development. What we need are heterogeneous model system with wide variability, if we want to develop drugs.
In other words, we sorely lack the training and the mindset that we need for drug development, because we’re trained to advance science, not to engineer solutions. And to top it off, scientists tend to look down on engineers, mainly because the engineers are not doing “real science.” When I was an undergraduate, Harvard, the archetypical liberal arts college, was in the process of eliminating the engineering department (shortly after it eliminated all accounting courses), presumably because it was too practical. Along with the bias against inventors, this bias against engineers tends to have very pernicious effect on the drug development field.
And how do engineers think? How do they differ? Well, here are some observations from “Applied Minds: How Engineers Think” by Madhavan.
“The core of the engineering mind-set is… modular systems thinking.… Systems level thinking is… about understanding that in the ebb and flow of life, nothing is stationary and everything is linked. The relationship among the modules of a system give rise to a whole that cannot be understood by analyzing its constituent parts.”
In other words, they understand emergent phenomena. It’s the direct opposite of how scientists think–we’re trained to break down everything into smallest piece possible, and not trained very much on how to put the piece back together.
As I’ve said before, while “in science one can learn the most by studying the least” (Marvin Minsky) one can understand the least by studying the least as well.
He also write that there are three essential properties of an engineering mind-set. The first is
“A structured systems-level thinking process would consider how the elements of the system are linked in logic in time, in sequence, and in function–and under what conditions they work and don’t work.”
Scientists do think about how things are linked, but not in a system-wide fashion. And we like to extrapolate from one reductionist instance to all conditions. For example, we often study proteins in a test tube, isolated from the cell, and alone by itself instead of in soup with all the other proteins. Then we extrapolate that the binding, activity, and other characteristics carry over into the cell.
But they don’t. For example, the specificity of kinases are not driven just by its structure but also by where exactly in the cell it is. The kinases are often exquisitely segregated into a sub-organelle of the cell, and achieve additional specificity because of where they are located. There are specific kinase receptors that tether kinases to very precise locations in the cell. Similarly, lymphocytes are programmed in the lymph node by migrating to a specific part of the node.
Similarly, we believe that mutations cause cancer, but now that we’re doing sequencing survey studies, it’s clear that many cells with what should be devastating oncogene mutations at completely benign. In almost every person, there are cells all over our skin that have tumor mutations, that are acting completely benign, It’c clear that tumor mutations are necessary but not sufficient cause of tumors,
You can’t just take a protein, a gene, or signal molecule out of context, characterize it, and then expect it to work the same way in vivo.
Madhavan also writes,
“The second attribute of the engineering mind-set is the adeptness at designing under constraints.… Constrains don’t permit engineers to wait until all phenomena are fully understood and explained.”
Beautifully stated. Perhaps the most important part of this quote is that engineers recognize that they’re working under conditions that they don’t fully understand. And they make allowances for it. For example, they build models of a bridge before they actually construct it. And they double everything in case they’re wrong about some of their assumptions.
This is not what we do as scientists. We’re trained to explain the world the best we can, and to use that theory as if it were real. We are not trained to operate as if we’re missing some data. If we think a kinase controls cell growth, we’re trained to expect that blocking that kinase will slow down the growth. And we’re surprised when it doesn’t.
The third attribute is trade-offs, but that doesn’t distinguish scientists and engineers.
But here is a quote that speaks to the heart of the matter.
“If the core of science is discovery, then the essence of engineering is creation.”
A different way of putting it may be that scientists are trained to build the best possible mental model of the facts and data points we know about, while the engineers are trained to build best possible real things in the real world, while accounting for the unknown unknowns.
So what can we do? We need to move even further away from target-based discovery. It doesn’t work, at least very often and certainly not yet. One day, we may understand enough about biology to do target-based discovery, but as of now, it’s just a conceit if we think we are really doing target-based rational drug discovery. Even the example held up a a shining example of target therapy, BRAF inhibition, is more of an accident than true rational drug design, in my opinion. Sure, it worked exactly as expected in melanoma–phenomenally, but it work exactly opposite of expected in colorectal cancer–complete failure. It ain’t target-based therapy if it don’t work on the target, you know?
Perhaps the best way to state this is: we as scientists need to be more humble. We need to be more pragmatic and empiric. I will leave you with two quotes.
“In theory there is no difference betwen theory and practice but in practice there is.”
“Americans ask, sure it works in theory but does it work in practice and Frenchmen ask, sure it works in practice but does it work in theory.”